feat: migrate Bedrock provider to AI SDK (#11243)

* feat: migrate Bedrock provider to AI SDK

Replace the raw AWS SDK (@aws-sdk/client-bedrock-runtime) Bedrock handler
with the Vercel AI SDK (@ai-sdk/amazon-bedrock). Reduces provider from
1,633 lines to 575 lines (65% reduction).

Key changes:
- Use streamText()/generateText() instead of ConverseStreamCommand/ConverseCommand
- Use createAmazonBedrock() with native auth (access key, secret, session,
  profile via credentialProvider, API key, VPC endpoint as baseURL)
- Reasoning config via providerOptions.bedrock.reasoningConfig
- Anthropic beta headers via providerOptions.bedrock.anthropicBeta
- Thinking signature captured from providerMetadata.bedrock.signature
  on reasoning-delta stream events
- Thinking signature round-tripped via providerOptions.bedrock.signature
  on reasoning parts in convertToAiSdkMessages()
- Redacted thinking captured from providerMetadata.bedrock.redactedData
- isAiSdkProvider() returns true for reasoning block preservation
- Keep: getModel, ARN parsing, cross-region inference, cost calculation,
  service tier pricing, 1M context beta

Tests: 83 tests skipped (mock old AWS SDK internals, need rewrite for
AI SDK mocking). 106 tests pass. 0 tests fail.

* fix: address review feedback for Bedrock AI SDK migration

- Wire usePromptCache into AI SDK via providerOptions.bedrock.cachePoint
  on system prompt and last two user messages
- Remove debug logger.info that fires on every stream event with
  providerMetadata
- Tighten isThrottlingError to match 'rate limit' instead of broad
  'rate'/'limit' substrings that false-positive on context length errors
- Use shared handleAiSdkError utility for consistent error handling
  with status code preservation for retry logic

* fix: bedrock AI SDK migration - fix usage metrics, rewrite tests, remove dead code

- Fix reasoningTokens always 0 (usage.details?.reasoningTokens → usage.reasoningTokens)
- Fix cacheReadInputTokens always 0 (read from usage.inputTokenDetails instead of providerMetadata)
- Fix invokedModelId not extracted for prompt router cost calculation
- Rewrite all 6 skipped bedrock test suites for AI SDK mocking pattern (140 tests pass)
- Remove dead code: bedrock-converse-format.ts, cache-strategy/ (6 files, ~2700 lines)

* chore: remove dead @anthropic-ai/bedrock-sdk dep and stale AWS SDK mocks

* chore: update pnpm-lock.yaml after removing @anthropic-ai/bedrock-sdk

* fix: compute cache point indices from original Anthropic messages before AI SDK conversion

The previous approach naively targeted the last 2 user messages in the
post-conversion AI SDK array, but convertToAiSdkMessages() splits user
messages containing tool_results into separate tool + user messages,
causing cache points to land on the wrong messages (tiny text fragments
instead of the intended meaty user turns).

Now we identify the last 2 user messages in the original Anthropic
message array (matching the Anthropic provider's caching strategy) and
build a parallel-walk mapping to apply cachePoint to the correct
corresponding AI SDK message.

* perf: optimize prompt caching with 3-point message strategy + anchor for 20-block window

Previous approach only cached the last 2 user messages (using 2 of 4
available cache checkpoints for messages). This left significant cache
savings on the table for longer conversations.

New strategy uses up to 3 message cache points (+ 1 system = 4 total):
- Last user message: write to cache for next request
- Second-to-last user message: read from cache for current request
- Anchor message at ~1/3 position: ensures the 20-block lookback window
  from the second-to-last breakpoint hits a stable cache entry, covering
  all assistant/tool messages in the middle of the conversation

Also extracted the parallel-walk mapping logic into a reusable
applyCachePointsToAiSdkMessages() helper method.

Industry benchmarks show 70-95% token cache rates are achievable;
this change should significantly improve our 39% baseline for longer
multi-turn conversations.

* chore: remove stale bedrock-sdk external, fix arnInfo property name, remove unused exports

---------

Co-authored-by: daniel-lxs <[email protected]>

apps/cli/tsup.config.ts

@@ -16,7 +16,6 @@ export default defineConfig({
 	external: [
 		// Keep native modules external
 		"@anthropic-ai/sdk",
-		"@anthropic-ai/bedrock-sdk",
 		"@anthropic-ai/vertex-sdk",
 		// Keep @vscode/ripgrep external - we bundle the binary separately
 		"@vscode/ripgrep",

pnpm-lock.yaml

@@ -746,6 +746,9 @@ importers:
 
   src:
     dependencies:
+      '@ai-sdk/amazon-bedrock':
+        specifier: ^4.0.50
+        version: 4.0.50([email protected])
       '@ai-sdk/cerebras':
         specifier: ^1.0.0
         version: 1.0.35([email protected])
@@ -770,9 +773,6 @@ importers:
       '@ai-sdk/xai':
         specifier: ^3.0.46
         version: 3.0.46([email protected])
-      '@anthropic-ai/bedrock-sdk':
-        specifier: ^0.10.2
-        version: 0.10.4
       '@anthropic-ai/sdk':
         specifier: ^0.37.0
         version: 0.37.0
@@ -1417,12 +1417,24 @@ packages:
   '@adobe/[email protected]':
     resolution: {integrity: sha512-baYZExFpsdkBNuvGKTKWCwKH57HRZLVtycZS05WTQNVOiXVSeAki3nU35zlRbToeMW8aHlJfyS+1C4BOv27q0A==}
 
+  '@ai-sdk/[email protected]':
+    resolution: {integrity: sha512-DsIxaUHPbDUY0DfxYMz6GL9tO/z7ISiwACSiYupcYImqrcdLtIGFujPgszOf92ed3olfhjdkhTwKBHaf6Yh6Qw==}
+    engines: {node: '>=18'}
+    peerDependencies:
+      zod: 3.25.76
+
   '@ai-sdk/[email protected]':
     resolution: {integrity: sha512-CkNW5L1Arv8gPtPlEmKd+yf/SG9ucJf0XQdpMG8OiYEtEMc2smuCA+tyCp8zI7IBVg/FE7nUfFHntQFaOjRwJQ==}
     engines: {node: '>=18'}
     peerDependencies:
       zod: 3.25.76
 
+  '@ai-sdk/[email protected]':
+    resolution: {integrity: sha512-tEgcJPw+a6obbF+SHrEiZsx3DNxOHqeY8bK4IpiNsZ8YPZD141R34g3lEAaQnmNN5mGsEJ8SXoEDabuzi8wFJQ==}
+    engines: {node: '>=18'}
+    peerDependencies:
+      zod: 3.25.76
+
   '@ai-sdk/[email protected]':
     resolution: {integrity: sha512-JrNdMYptrOUjNthibgBeAcBjZ/H+fXb49sSrWhOx5Aq8eUcrYvwQ2DtSAi8VraHssZu78NAnBMrgFWSUOTXFxw==}
     engines: {node: '>=18'}
@@ -1575,9 +1587,6 @@ packages:
   '@antfu/[email protected]':
     resolution: {integrity: sha512-Mex9nXf9vR6AhcXmMrlz/HVgYYZpVGJ6YlPgwl7UnaFpnshXs6EK/oa5Gpf3CzENMjkvEx2tQtntGnb7UtSTOQ==}
 
-  '@anthropic-ai/[email protected]':
-    resolution: {integrity: sha512-szduEHbMli6XL934xrraYg5cFuKL/1oMyj/iZuEVjtddQ7eD5cXObzWobsv5mTLWijQmSzMfFD+JAUHDPHlQ/Q==}
-
   '@anthropic-ai/[email protected]':
     resolution: {integrity: sha512-tHjX2YbkUBwEgg0JZU3EFSSAQPoK4qQR/NFYa8Vtzd5UAyXzZksCw2In69Rml4R/TyHPBfRYaLK35XiOe33pjw==}
 
@@ -1587,9 +1596,6 @@ packages:
   '@asamuzakjp/[email protected]':
     resolution: {integrity: sha512-K1A6z8tS3XsmCMM86xoWdn7Fkdn9m6RSVtocUrJYIwZnFVkng/PvkEoWtOWmP+Scc6saYWHWZYbndEEXxl24jw==}
 
-  '@aws-crypto/[email protected]':
-    resolution: {integrity: sha512-IzSgsrxUcsrejQbPVilIKy16kAT52EwB6zSaI+M3xxIhKh5+aldEyvI+z6erM7TCLB2BJsFrtHjp6/4/sr+3dA==}
-
   '@aws-crypto/[email protected]':
     resolution: {integrity: sha512-nLbCWqQNgUiwwtFsen1AdzAtvuLRsQS8rYgMuxCrdKf9kOssamGLuPwyTY9wyYblNr9+1XM8v6zoDTPPSIeANg==}
     engines: {node: '>=16.0.0'}
@@ -1597,9 +1603,6 @@ packages:
   '@aws-crypto/[email protected]':
     resolution: {integrity: sha512-AXfN/lGotSQwu6HNcEsIASo7kWXZ5HYWvfOmSNKDsEqC4OashTp8alTmaz+F7TC2L083SFv5RdB+qU3Vs1kZqw==}
 
-  '@aws-crypto/[email protected]':
-    resolution: {integrity: sha512-MHGJyjE7TX9aaqXj7zk2ppnFUOhaDs5sP+HtNS0evOxn72c+5njUmyJmpGd7TfyoDznZlHMmdo/xGUdu2NIjNQ==}
-
   '@aws-crypto/[email protected]':
     resolution: {integrity: sha512-FFQQyu7edu4ufvIZ+OadFpHHOt+eSTBaYaki44c+akjg7qZg9oOQeLlk77F6tSYqjDAFClrHJk9tMf0HdVyOvA==}
     engines: {node: '>=16.0.0'}
@@ -1607,12 +1610,6 @@ packages:
   '@aws-crypto/[email protected]':
     resolution: {integrity: sha512-iAvUotm021kM33eCdNfwIN//F77/IADDSs58i+MDaOqFrVjZo9bAal0NK7HurRuWLLpF1iLX7gbWrjHjeo+YFg==}
 
-  '@aws-crypto/[email protected]':
-    resolution: {integrity: sha512-2OJlpeJpCR48CC8r+uKVChzs9Iungj9wkZrl8Z041DWEWvyIHILYKCPNzJghKsivj+S3mLo6BVc7mBNzdxA46w==}
-
-  '@aws-crypto/[email protected]':
-    resolution: {integrity: sha512-2EnmPy2gsFZ6m8bwUQN4jq+IyXV3quHAcwPOS6ZA3k+geujiqI8aRokO2kFJe+idJ/P3v4qWI186rVMo0+zLDQ==}
-
   '@aws-crypto/[email protected]':
     resolution: {integrity: sha512-4RkU9EsI6ZpBve5fseQlGNUWKMa1RLPQ1dnjnQoe07ldfIzcsGb5hC5W0Dm7u423KWzawlrpbjXBrXCEv9zazQ==}
 
@@ -1708,14 +1705,6 @@ packages:
     resolution: {integrity: sha512-/inmPnjZE0ZBE16zaCowAvouSx05FJ7p6BQYuzlJ8vxEU0sS0Hf8fvhuiRnN9V9eDUPIBY+/5EjbMWygXL4wlQ==}
     engines: {node: '>=18.0.0'}
 
-  '@aws-sdk/[email protected]':
-    resolution: {integrity: sha512-A9qnsy9zQ8G89vrPPlNG9d1d8QcKRGqJKqwyGgS0dclJpwy6d1EWgQLIolKPl6vcFpLoe6avLOLxr+h8ur5wpg==}
-    engines: {node: '>=18.0.0'}
-
-  '@aws-sdk/[email protected]':
-    resolution: {integrity: sha512-xliuHaUFZxEx1NSXeLLZ9Dyu6+EJVQKEoD+yM+zqUo3YDZ7medKJWY6fIOKiPX/N7XbLdBYwajb15Q7IL8KkeA==}
-    engines: {node: '>=18.0.0'}
-
   '@aws-sdk/[email protected]':
     resolution: {integrity: sha512-eLA6XjVobAUAMivvM7DBL79mnHyrm+32TkXNWZua5mnxF+6kQCfblKKJvxMZLGosO53/Ex46ogim8IY5Nbqv2w==}
     engines: {node: '>=18.0.0'}
@@ -1744,9 +1733,6 @@ packages:
       aws-crt:
         optional: true
 
-  '@aws-sdk/[email protected]':
-    resolution: {integrity: sha512-UvFa/vR+e19XookZF8RzFZBrw2EUkQWxiBW0yYQAhvk3C+QVGl0H3ouca8LDBlBfQKXwmW3huo/59H8rwb1wJw==}
-
   '@aws-sdk/[email protected]':
     resolution: {integrity: sha512-LVHg0jgjyicKKvpNIEMXIMr1EBViESxcPkqfOlT+X1FkmUMTNZEEVF18tOJg4m4hV5vxtkWcqtr4IEeWa1C41Q==}
     engines: {node: '>=18.0.0'}
@@ -3879,10 +3865,6 @@ packages:
     resolution: {integrity: sha512-tlqY9xq5ukxTUZBmoOp+m61cqwQD5pHJtFY3Mn8CA8ps6yghLH/Hw8UPdqg4OLmFW3IFlcXnQNmo/dh8HzXYIQ==}
     engines: {node: '>=18'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-wRlta7GuLWpTqtFfGo+nZyOO1vEvewdNR1R4rTxpC8XU6vG/NDyrFBhwLZsqg1NUoR1noVaXJPC/7ZK47QCySw==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-Z4DUr/AkgyFf1bOThW2HwzREagee0sB5ycl+hDiSZOfRLW8ZgrOjDi6g8mHH19yyU5E2A/64W3z6SMIf5XiUSQ==}
     engines: {node: '>=18.0.0'}
@@ -3899,9 +3881,6 @@ packages:
     resolution: {integrity: sha512-YVNMjhdz2pVto5bRdux7GMs0x1m0Afz3OcQy/4Yf9DH4fWOtroGH7uLvs7ZmDyoBJzLdegtIPpXrpJOZWvUXdw==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-8janZoJw85nJmQZc4L8TuePp2pk1nxLgkxIR0TUjKJ5Dkj5oelB9WtiSSGXCQvNsJl0VSTvK/2ueMXxvpa9GVw==}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-aV8blR9RBDKrOlZVgjOdmOibTC2sBXNiT7WA558b4MPdsLTV6sbyc1WIE9QiIuYMJjYtnPLciefoqSW8Gi+MZQ==}
     engines: {node: '>=18.0.0'}
@@ -3914,25 +3893,14 @@ packages:
     resolution: {integrity: sha512-lxfDT0UuSc1HqltOGsTEAlZ6H29gpfDSdEPTapD5G63RbnYToZ+ezjzdonCCH90j5tRRCw3aLXVbiZaBW3VRVg==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-zpQMtJVqCUMn+pCSFcl9K/RPNtQE0NuMh8sKpCdEHafhwRsjP50Oq/4kMmvxSRy6d8Jslqd8BLvDngrUtmN9iA==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-TPhiGByWnYyzcpU/K3pO5V7QgtXYpE0NaJPEZBCa1Y5jlw5SjqzMSbFiLb+ZkJhqoQc0ImGyVINqnq1ze0ZRcQ==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-pvoe/vvJY0mOpuF84BEtyZoYfbehiFj8KKWk1ds2AT0mTLYFVs+7sBJZmioOFdBXKd48lfrx1vumdPdmGlCLxA==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-GNI/IXaY/XBB1SkGBFmbW033uWA0tj085eCxYih0eccUe/PFR7+UBQv9HNDk2fD9TJu7UVsCWsH99TkpEPSOzQ==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-BOWEBeppWhLn/no/JxUL/ghTfANTjT7kg3Ww2rPqTUY9R4yHPXxJ9JhMe3Z03LN3aPwiwlpDIUcVw1xDyHqEhw==}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-mg83SM3FLI8Sa2ooTJbsh5MFfyMTyNRwxqpKHmE0ICRIa66Aodv80DMsTQI02xBLVJ0hckwqTRr5IGAbbWuFLQ==}
     engines: {node: '>=18.0.0'}
@@ -3949,10 +3917,6 @@ packages:
     resolution: {integrity: sha512-GGP3O9QFD24uGeAXYUjwSTXARoqpZykHadOmA8G5vfJPK0/DC67qa//0qvqrJzL1xc8WQWX7/yc7fwudjPHPhA==}
     engines: {node: '>=14.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-+Fsu6Q6C4RSJiy81Y8eApjEB5gVtM+oFKTffg+jSuwtvomJJrhUJBu2zS8wjXSgH/g1MKEWrzyChTBe6clb5FQ==}
-    engines: {node: '>=16.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-DZZZBvC7sjcYh4MazJSGiWMI2L7E0oCiRHREDzIxi/M2LY79/21iXt6aPLHge82wi5LsuRF5A06Ds3+0mlh6CQ==}
     engines: {node: '>=18.0.0'}
@@ -3961,10 +3925,6 @@ packages:
     resolution: {integrity: sha512-hJRZuFS9UsElX4DJSJfoX4M1qXRH+VFiLMUnhsWvtOOUWRNvvOfDaUSdlNbjwv1IkpVjj/Rd/O59Jl3nhAcxow==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-1/8kFp6Fl4OsSIVTWHnNjLnTL8IqpIb/D3sTSczrKFnrE9VMNWxnrRKNvpUHOJ6zpGD5f62TPm7+17ilTJpiCQ==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-PXehXofGMFpDqr933rxD8RGOcZ0QBAWtuzTgYRAHAL2BnKawHDEdf/TnGpcmfPJGwonhginaaeJIKluEojiF/w==}
     engines: {node: '>=18.0.0'}
@@ -3973,66 +3933,34 @@ packages:
     resolution: {integrity: sha512-OhLx131znrEDxZPAvH/OYufR9d1nB2CQADyYFN4C3V/NQS7Mg4V6uvxHC/Dr96ZQW8IlHJTJ+vAhKt6oxWRndA==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-sIADe7ojwqTyvEQBe1nc/GXB9wdHhi9UwyX0lTyttmUWDJLP655ZYE1WngnNyXREme8I27KCaUhyhZWRXL0q7Q==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-jUr3x2CDhV15TOX2/Uoz4gfgeqLrRoTQbYAuhLS7lcVKNev7FeYSJ1ebEfjk+l9kbb7k7LfzIR/irgxys5ZTOg==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-Qntc3jrtwwrsAC+X8wms8zhrTr0sFXnyEGhZd9sLtsJ/6gGQKFzNB+wWbOcpJd7BR8ThNCoKt76BuQahfMvpeA==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-Gy3TKCOnm9JwpFooldwAboazw+EFYlC+Bb+1QBsSi5xI0W5lX81j/P5+CXvD/9ZjtYKRgxq+kkqd/KOHflzvgA==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-0elK5/03a1JPWMDPaS726Iw6LpQg80gFut1tNpPfxFuChEEklo2yL823V94SpTZTxmKlXFtFgsP55uh3dErnIg==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-3X3w7qzmo4XNNdPKNS4nbJcGSwiEMsNsRSunMA92S4DJLLIrH5g1AyuOA2XKM9PAPi8mIWfqC+fnfKNsI4KvHw==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-mVGyPBzkkGQsPoxQUbxlEfRjrj6FPyA3u3u2VXGr9hT8wilsoQdZdvKpMBFMB8Crfhv5dNkKHIW0Yyuc7eABqA==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-VXHGfzCXLZeKnFp6QXjAdy+U8JF9etfpUXD1FAbzY1GzsFJiDQRQIt2CnMUvUdz3/YaHNqT3RphVWMUpXTIODA==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-+xiil2lFhtTRzXkx8F053AV46QnIw6e7MV8od5Mi68E1ICOjCeCHw2XfLnDEUHnT9WGUIkwcqavXjfwuJbGlpg==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-g2DHo08IhxV5GdY3Cpt/jr0mkTlAD39EJKN27Jb5N8Fb5qt8KG39wVKTXiTRCmHHou7lbXR8nKVU14/aRUf86w==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-Xy5XK1AFWW2nlY/biWZXu6/krgbaf2dg0q492D8M5qthsnU2H+UgFeZLbM76FnH7s6RO/xhQRkj+T6KBO3JzgQ==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-3sfFd2MAzVt0Q/klOmjFi3oIkxczHs0avbwrfn1aBqtc23WqQSmjvk77MBw9WkEQcwbOYIX5/2z4ULj8DuxSsw==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-L1kSeviUWL+emq3CUVSgdogoM/D9QMFaqxL/dd0X7PCNWmPXqt+ExtrBjqT0V7HLN03Vs9SuiLrG3zy3JGnE5A==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-KQ1gFXXC+WsbPFnk7pzskzOpn4s+KheWgO3dzkIEmnb6NskAIGp/dGdbKisTPJdtov28qNDohQrgDUKzXZBLig==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-BvHCDrKfbG5Yhbpj4vsbuPV2GgcpHiAkLeIlcA1LtfpMz3jrqizP1+OguSNSj1MwBHEiN+jwNisXLGdajGDQJA==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-aHb5cqXZocdzEkZ/CvhVjdw5l4r1aU/9iMEyoKzH4eXMowT6M0YjBpp7W/+XjkBnY8Xh0kVd55GKjnPKlCwinQ==}
     engines: {node: '>=18.0.0'}
@@ -4041,57 +3969,26 @@ packages:
     resolution: {integrity: sha512-fdWuhEx4+jHLGeew9/IvqVU/fxT/ot70tpRGuOLxE3HzZOyKeTQfYeV1oaBXpzi93WOk668hjMuuagJ2/Qs7ng==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-WyujUJL8e1B6Z4PBfAqC/aGY1+C7T0w20Gih3yrvJSk97gpiVfB+y7c46T4Nunk+ZngLq0rOIdeVeIklk0R3OA==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-y5ozxeQ9omVjbnJo9dtTsdXj9BEvGx2X8xvRgKnV+/7wLBuYJQL6dOa/qMY6omyHi7yjt1OA97jZLoVRYi8lxA==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-3BcPylEsYtD0esM4Hoyml/+s7WP2LFhcM3J2AGdcL2vx9O60TtfpDOL72gjb4lU8NeRPeKAwR77YNyyGvMbuEA==}
-    engines: {node: '>=16.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-ScDCpasxH7w1HXHYbtk3jcivjvdA1VICyAdgvVqKhKKwxi+MTwZEqFw0minE+oZ7F07oF25xh4FGJxgqgShz0A==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-jrbSQrYCho0yDaaf92qWgd+7nAeap5LtHTI51KXqmpIFCceKU3K9+vIVTUH72bOJngBMqa4kyu1VJhRcSrk/CQ==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-gZU4uAFcdrSi3io8U99Qs/FvVdRxPvIMToi+MFfsy/DN9UqtknJ1ais+2M9yR8e0ASQpNmFYEKeIKVcMjQg3rg==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-QwYgloJ0sVNBeBuBs65cIkTbfzV/Q6ZNPCJ99EICFEdJYG50nGIY/uYXp+TbsdJReIuPr0a0kXmCvren3MbRRw==}
-    engines: {node: '>=14.0.0'}
-
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-bNwBYYmN8Eh9RyjS1p2gW6MIhSO2rl7X9QeLM8iTdcGRP+eDiIWDt66c9IysCc22gefKszZv+ubV9qZc7hdESg==}
-    engines: {node: '>=16.0.0'}
-
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-UqKOQBL2x6+HWl3P+3QqFD4ncKq0I8Nuz9QItGv5WuKuMHuuwlhvqcZCoXGfc+P1QmfJE7VieykoYYmrOoFJxA==}
-    engines: {node: '>=18.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-N0Zn0OT1zc+NA+UVfkYqQzviRh5ucWwO7mBV3TmHHprMnfcJNfhlPicDkBHi0ewbh+y3evR6cNAW0Raxvb01NA==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-hoA4zm61q1mNTpksiSWp2nEl1dt3j726HdRhiNgVJQMj7mLp7dprtF57mOB6JvEk/x9d2bsuL5hlqZbBuHQylQ==}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-w/N/Iw0/PTwJ36PDqU9PzAwVElo4qXxCC0eCTlUtIz/Z5V/2j/cViMHi0hPukSBHp4DVwvUlUhLgCzqSJ6plrg==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-s3+eVwNeJuXUwuMbusncZNViuhv2LjVJ1nMwTqSA0XAC7gjKhqqxRdJPhR8+YrkoZ9IiIbFk/yK6ACe/xlF+hw==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-GkXZ59JfyxsIwNTWFnjmFEI8kZpRNIBfxKjv09+nkAWPt/4aGaEWMM04m4sxgNVWkbt2MdSvE3KF/PfX4nFedQ==}
     engines: {node: '>=18.0.0'}
@@ -4108,10 +4005,6 @@ packages:
     resolution: {integrity: sha512-IJdWBbTcMQ6DA0gdNhh/BwrLkDR+ADW5Kr1aZmd4k3DIF6ezMV4R2NIAmT08wQJ3yUK82thHWmC/TnK/wpMMIA==}
     engines: {node: '>=14.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-aEOHCgq5RWFbP+UDPvPot26EJHjOC+bRgse5A8V3FSShqd5E5UN4qc7zkwsvJPPAVsf73QwYcHN1/gt/rtLwQA==}
-    engines: {node: '>=16.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-kAY9hTKulTNevM2nlRtxAG2FQ3B2OR6QIrPY3zE5LqJy1oxzmgBGsHLWTcNhWXKchgA0WHW+mZkQrng/pgcCew==}
     engines: {node: '>=18.0.0'}
@@ -4132,26 +4025,10 @@ packages:
     resolution: {integrity: sha512-f+nBDhgYRCmUEDKEQb6q0aCcOTXRDqH5wWaFHJxt4anB4pKHlgGoYP3xtioKXH64e37ANUkzWf6p4Mnv1M5/Vg==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-7iKXR+/4TpLK194pVjKiasIyqMtTYJsgKgM242Y9uzt5dhHnUDvMNb+3xIhRJ9QhvqGii/5cRUt4fJn3dtXNHQ==}
-    engines: {node: '>=14.0.0'}
-
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-eFndh1WEK5YMUYvy3lPlVmYY/fZcQE1D8oSf41Id2vCeIkKJXPcYDCZD+4+xViI6b1XSd7tE+s5AmXzz5ilabQ==}
-    engines: {node: '>=16.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-CCQBwJIvXMLKxVbO88IukazJD9a4kQ9ZN7/UMGBjBcJYvatpWk+9g870El4cB8/EJxfe+k+y0GmR9CAzkF+Nbw==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-L1qpleXf9QD6LwLCJ5jddGkgWyuSvWBkJwWAZ6kFkdifdso+sk3L3O1HdmPvCdnCK3IS4qWyPxev01QMnfHSBw==}
-    engines: {node: '>=14.0.0'}
-
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-dWpyc1e1R6VoXrwLoLDd57U1z6CwNSdkM69Ie4+6uYh2GC7Vg51Qtan7ITzczuVpqezdDTKJGJB95fFvvjU/ow==}
-    engines: {node: '>=16.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-fKGQAPAn8sgV0plRikRVo6g6aR0KyKvgzNrPuM74RZKy/wWVzx3BMk+ZWEueyN3L5v5EDg+P582mKU+sH5OAsg==}
     engines: {node: '>=18.0.0'}
@@ -4160,22 +4037,10 @@ packages:
     resolution: {integrity: sha512-yQncJmj4dtv/isTXxRb4AamZHy4QFr4ew8GxS6XLWt7sCIxkPxPzINWd7WLISEFPsIan14zrKgvyAF+/yzfwoA==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-17faEXbYWIRst1aU9SvPZyMdWmqIrduZjVOqCPMIsWFNxs5yQQgFrJL6b2SdiCzyW9mJoDjFtgi53xx7EH+BXA==}
-    engines: {node: '>=14.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-7M5aVFjT+HPilPOKbOmQfCIPchZe4DSBc1wf1+NvHvSoFTiFtauZzT+onZvCj70xhXd0AEmYnZYmdJIuwxOo4w==}
     engines: {node: '>=18.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-jtmJMyt1xMD/d8OtbVJ2gFZOSKc+ueYJZPW20ULW1GOp/q/YIM0wNh+u8ZFao9UaIGz4WoPW8hC64qlWLIfoDA==}
-    engines: {node: '>=14.0.0'}
-
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-LqR7qYLgZTD7nWLBecUi4aqolw8Mhza9ArpNEQ881MJJIU2sE5iHCK6TdyqqzcDLy0OPe10IY4T8ctVdtynubg==}
-    engines: {node: '>=16.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-igZpCKV9+E/Mzrpq6YacdTQ0qTiLm85gD6N/IrmyDvQFA4UnU3d5g3m8tMT/6zG/vVkWSU+VxeUyGonL62DuxA==}
     engines: {node: '>=18.0.0'}
@@ -4184,10 +4049,6 @@ packages:
     resolution: {integrity: sha512-R8Rdn8Hy72KKcebgLiv8jQcQkXoLMOGGv5uI1/k0l+snqkOzQ1R0ChUBCxWMlBsFMekWjq0wRudIweFs7sKT5A==}
     engines: {node: '>=14.0.0'}
 
-  '@smithy/[email protected]':
-    resolution: {integrity: sha512-rUeT12bxFnplYDe815GXbq/oixEGHfRFFtcTF3YdDi/JaENIM6aSYYLJydG83UNzLXeRI5K8abYd/8Sp/QM0kA==}
-    engines: {node: '>=16.0.0'}
-
   '@smithy/[email protected]':
     resolution: {integrity: sha512-zBPfuzoI8xyBtR2P6WQj63Rz8i3AmfAaJLuNG8dWsfvPe8lO4aCPYLn879mEgHndZH1zQ2oXmG8O1GGzzaoZiw==}
     engines: {node: '>=18.0.0'}
@@ -5146,6 +5007,9 @@ packages:
     resolution: {integrity: sha512-wvUjBtSGN7+7SjNpq/9M2Tg350UZD3q62IFZLbRAR1bSMlCo1ZaeW+BJ+D090e4hIIZLBcTDWe4Mh4jvUDajzQ==}
     engines: {node: '>= 0.4'}
 
+  [email protected]:
+    resolution: {integrity: sha512-/djoAN709iY65ETD6LKCtyyEI04XIBP5xVvfmNxsEP0uJB5tyaGBztSryRr4HqMStr9R06PisQE7m9zDTXKu6g==}
+
   [email protected]:
     resolution: {integrity: sha512-oXTDccv8PcfjZmPGlWsPSwtOJCZ/b6W5jAMCNcfwJbCzDckwG0jrYJFaWH1yvivfCXjVzV/SPDEhMB3Q+DSurg==}
 
@@ -11130,12 +10994,28 @@ snapshots:
 
   '@adobe/[email protected]': {}
 
+  '@ai-sdk/[email protected]([email protected])':
+    dependencies:
+      '@ai-sdk/anthropic': 3.0.37([email protected])
+      '@ai-sdk/provider': 3.0.7
+      '@ai-sdk/provider-utils': 4.0.13([email protected])
+      '@smithy/eventstream-codec': 4.2.4
+      '@smithy/util-utf8': 4.2.0
+      aws4fetch: 1.0.20
+      zod: 3.25.76
+
   '@ai-sdk/[email protected]([email protected])':
     dependencies:
       '@ai-sdk/provider': 2.0.1
       '@ai-sdk/provider-utils': 3.0.20([email protected])
       zod: 3.25.76
 
+  '@ai-sdk/[email protected]([email protected])':
+    dependencies:
+      '@ai-sdk/provider': 3.0.7
+      '@ai-sdk/provider-utils': 4.0.13([email protected])
+      zod: 3.25.76
+
   '@ai-sdk/[email protected]([email protected])':
     dependencies:
       '@ai-sdk/openai-compatible': 1.0.31([email protected])
@@ -11304,23 +11184,6 @@ snapshots:
 
   '@antfu/[email protected]': {}
 
-  '@anthropic-ai/[email protected]':
-    dependencies:
-      '@anthropic-ai/sdk': 0.37.0
-      '@aws-crypto/sha256-js': 4.0.0
-      '@aws-sdk/client-bedrock-runtime': 3.922.0
-      '@aws-sdk/credential-providers': 3.922.0
-      '@smithy/eventstream-serde-node': 2.2.0
-      '@smithy/fetch-http-handler': 2.5.0
-      '@smithy/protocol-http': 3.3.0
-      '@smithy/signature-v4': 3.1.2
-      '@smithy/smithy-client': 2.5.1
-      '@smithy/types': 2.12.0
-      '@smithy/util-base64': 2.3.0
-    transitivePeerDependencies:
-      - aws-crt
-      - encoding
-
   '@anthropic-ai/[email protected]':
     dependencies:
       '@types/node': 18.19.100
@@ -11349,12 +11212,6 @@ snapshots:
       '@csstools/css-tokenizer': 3.0.4
       lru-cache: 10.4.3
 
-  '@aws-crypto/[email protected]':
-    dependencies:
-      '@aws-crypto/util': 3.0.0
-      '@aws-sdk/types': 3.840.0
-      tslib: 1.14.1
-
   '@aws-crypto/[email protected]':
     dependencies:
       '@aws-crypto/util': 5.2.0
@@ -11371,12 +11228,6 @@ snapshots:
       '@smithy/util-utf8': 2.3.0
       tslib: 2.8.1
 
-  '@aws-crypto/[email protected]':
-    dependencies:
-      '@aws-crypto/util': 4.0.0
-      '@aws-sdk/types': 3.804.0
-      tslib: 1.14.1
-
   '@aws-crypto/[email protected]':
     dependencies:
       '@aws-crypto/util': 5.2.0
@@ -11387,18 +11238,6 @@ snapshots:
     dependencies:
       tslib: 2.8.1
 
-  '@aws-crypto/[email protected]':
-    dependencies:
-      '@aws-sdk/types': 3.840.0
-      '@aws-sdk/util-utf8-browser': 3.259.0
-      tslib: 1.14.1
-
-  '@aws-crypto/[email protected]':
-    dependencies:
-      '@aws-sdk/types': 3.840.0
-      '@aws-sdk/util-utf8-browser': 3.259.0
-      tslib: 1.14.1
-
   '@aws-crypto/[email protected]':
     dependencies:
       '@aws-sdk/types': 3.922.0
@@ -11806,16 +11645,6 @@ snapshots:
     transitivePeerDependencies:
       - aws-crt
 
-  '@aws-sdk/[email protected]':
-    dependencies:
-      '@smithy/types': 4.3.1
-      tslib: 2.8.1
-
-  '@aws-sdk/[email protected]':
-    dependencies:
-      '@smithy/types': 4.3.1
-      tslib: 2.8.1
-
   '@aws-sdk/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
@@ -11855,10 +11684,6 @@ snapshots:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@aws-sdk/[email protected]':
-    dependencies:
-      tslib: 2.8.1
-
   '@aws-sdk/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
@@ -14041,11 +13866,6 @@ snapshots:
 
   '@sindresorhus/[email protected]': {}
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
@@ -14081,13 +13901,6 @@ snapshots:
       '@smithy/url-parser': 4.2.4
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@aws-crypto/crc32': 3.0.0
-      '@smithy/types': 2.12.0
-      '@smithy/util-hex-encoding': 2.2.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@aws-crypto/crc32': 5.2.0
@@ -14106,38 +13919,18 @@ snapshots:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/eventstream-serde-universal': 2.2.0
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/eventstream-serde-universal': 4.2.4
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/eventstream-codec': 2.2.0
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/eventstream-codec': 4.2.4
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/protocol-http': 3.3.0
-      '@smithy/querystring-builder': 2.2.0
-      '@smithy/types': 2.12.0
-      '@smithy/util-base64': 2.3.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/protocol-http': 5.3.4
@@ -14162,10 +13955,6 @@ snapshots:
     dependencies:
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       tslib: 2.8.1
@@ -14176,16 +13965,6 @@ snapshots:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/middleware-serde': 2.3.0
-      '@smithy/node-config-provider': 2.3.0
-      '@smithy/shared-ini-file-loader': 2.4.0
-      '@smithy/types': 2.12.0
-      '@smithy/url-parser': 2.2.0
-      '@smithy/util-middleware': 2.2.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/core': 3.17.2
@@ -14209,34 +13988,17 @@ snapshots:
       '@smithy/uuid': 1.1.0
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/protocol-http': 5.3.4
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/property-provider': 2.2.0
-      '@smithy/shared-ini-file-loader': 2.4.0
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/property-provider': 4.2.4
@@ -14244,14 +14006,6 @@ snapshots:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/abort-controller': 2.2.0
-      '@smithy/protocol-http': 3.3.0
-      '@smithy/querystring-builder': 2.2.0
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/abort-controller': 4.2.4
@@ -14260,43 +14014,22 @@ snapshots:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 2.12.0
-      '@smithy/util-uri-escape': 2.2.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
       '@smithy/util-uri-escape': 4.2.0
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
@@ -14306,26 +14039,11 @@ snapshots:
     dependencies:
       '@smithy/types': 4.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/is-array-buffer': 3.0.0
-      '@smithy/types': 3.7.2
-      '@smithy/util-hex-encoding': 3.0.0
-      '@smithy/util-middleware': 3.0.11
-      '@smithy/util-uri-escape': 3.0.0
-      '@smithy/util-utf8': 3.0.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/is-array-buffer': 4.2.0
@@ -14337,15 +14055,6 @@ snapshots:
       '@smithy/util-utf8': 4.2.0
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/middleware-endpoint': 2.5.1
-      '@smithy/middleware-stack': 2.2.0
-      '@smithy/protocol-http': 3.3.0
-      '@smithy/types': 2.12.0
-      '@smithy/util-stream': 2.2.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/core': 3.17.2
@@ -14356,40 +14065,16 @@ snapshots:
       '@smithy/util-stream': 4.5.5
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      tslib: 2.8.1
-
-  '@smithy/[email protected]':
-    dependencies:
-      tslib: 2.8.1
-
-  '@smithy/[email protected]':
-    dependencies:
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/querystring-parser': 2.2.0
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/querystring-parser': 4.2.4
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/util-buffer-from': 2.2.0
-      '@smithy/util-utf8': 2.3.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/util-buffer-from': 4.2.0
@@ -14409,11 +14094,6 @@ snapshots:
       '@smithy/is-array-buffer': 2.2.0
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/is-array-buffer': 3.0.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/is-array-buffer': 4.2.0
@@ -14446,28 +14126,10 @@ snapshots:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      tslib: 2.8.1
-
-  '@smithy/[email protected]':
-    dependencies:
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 2.12.0
-      tslib: 2.8.1
-
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/types': 3.7.2
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/types': 4.8.1
@@ -14479,17 +14141,6 @@ snapshots:
       '@smithy/types': 4.8.1
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/fetch-http-handler': 2.5.0
-      '@smithy/node-http-handler': 2.5.0
-      '@smithy/types': 2.12.0
-      '@smithy/util-base64': 2.3.0
-      '@smithy/util-buffer-from': 2.2.0
-      '@smithy/util-hex-encoding': 2.2.0
-      '@smithy/util-utf8': 2.3.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/fetch-http-handler': 5.3.5
@@ -14501,14 +14152,6 @@ snapshots:
       '@smithy/util-utf8': 4.2.0
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      tslib: 2.8.1
-
-  '@smithy/[email protected]':
-    dependencies:
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       tslib: 2.8.1
@@ -14518,11 +14161,6 @@ snapshots:
       '@smithy/util-buffer-from': 2.2.0
       tslib: 2.8.1
 
-  '@smithy/[email protected]':
-    dependencies:
-      '@smithy/util-buffer-from': 3.0.0
-      tslib: 2.8.1
-
   '@smithy/[email protected]':
     dependencies:
       '@smithy/util-buffer-from': 4.2.0
@@ -15627,6 +15265,8 @@ snapshots:
     dependencies:
       possible-typed-array-names: 1.1.0
 
+  [email protected]: {}
+
   [email protected]:
     dependencies:
       follow-redirects: 1.15.11

src/api/providers/__tests__/bedrock-custom-arn.spec.ts

@@ -22,38 +22,6 @@ vitest.mock("../../../utils/logging", () => ({
 	},
 }))
 
-// Mock AWS SDK
-vitest.mock("@aws-sdk/client-bedrock-runtime", () => {
-	const mockModule = {
-		lastCommandInput: null as Record<string, any> | null,
-		mockSend: vitest.fn().mockImplementation(async function () {
-			return {
-				output: new TextEncoder().encode(JSON.stringify({ content: "Test response" })),
-			}
-		}),
-		mockConverseCommand: vitest.fn(function (input) {
-			mockModule.lastCommandInput = input
-			return { input }
-		}),
-		MockBedrockRuntimeClient: class {
-			public config: any
-			public send: any
-
-			constructor(config: { region?: string }) {
-				this.config = config
-				this.send = mockModule.mockSend
-			}
-		},
-	}
-
-	return {
-		BedrockRuntimeClient: mockModule.MockBedrockRuntimeClient,
-		ConverseCommand: mockModule.mockConverseCommand,
-		ConverseStreamCommand: vitest.fn(),
-		__mock: mockModule, // Expose mock internals for testing
-	}
-})
-
 describe("Bedrock ARN Handling", () => {
 	// Helper function to create a handler with specific options
 	const createHandler = (options: Partial<ApiHandlerOptions> = {}) => {
@@ -224,8 +192,8 @@ describe("Bedrock ARN Handling", () => {
 					"arn:aws:bedrock:eu-west-1:123456789012:inference-profile/anthropic.claude-3-sonnet-20240229-v1:0",
 			})
 
-			// Verify the client was created with the ARN region, not the provided region
-			expect((handler as any).client.config.region).toBe("eu-west-1")
+			// Verify the handler's options were updated with the ARN region
+			expect((handler as any).options.awsRegion).toBe("eu-west-1")
 		})
 
 		it("should log region mismatch warning when ARN region differs from provided region", () => {

src/api/providers/__tests__/bedrock-error-handling.spec.ts

@@ -8,9 +8,6 @@ vi.mock("@roo-code/telemetry", () => ({
 	},
 }))
 
-// Mock BedrockRuntimeClient and commands
-const mockSend = vi.fn()
-
 // Mock AWS SDK credential providers
 vi.mock("@aws-sdk/credential-providers", () => {
 	return {
@@ -21,16 +18,27 @@ vi.mock("@aws-sdk/credential-providers", () => {
 	}
 })
 
-vi.mock("@aws-sdk/client-bedrock-runtime", () => ({
-	BedrockRuntimeClient: vi.fn().mockImplementation(() => ({
-		send: mockSend,
-	})),
-	ConverseStreamCommand: vi.fn(),
-	ConverseCommand: vi.fn(),
+// Use vi.hoisted to define mock functions for AI SDK
+const { mockStreamText, mockGenerateText } = vi.hoisted(() => ({
+	mockStreamText: vi.fn(),
+	mockGenerateText: vi.fn(),
+}))
+
+vi.mock("ai", async (importOriginal) => {
+	const actual = await importOriginal<typeof import("ai")>()
+	return {
+		...actual,
+		streamText: mockStreamText,
+		generateText: mockGenerateText,
+	}
+})
+
+vi.mock("@ai-sdk/amazon-bedrock", () => ({
+	createAmazonBedrock: vi.fn(() => vi.fn(() => ({ modelId: "test", provider: "bedrock" }))),
 }))
 
 import { AwsBedrockHandler } from "../bedrock"
-import { Anthropic } from "@anthropic-ai/sdk"
+import type { Anthropic } from "@anthropic-ai/sdk"
 
 describe("AwsBedrockHandler Error Handling", () => {
 	let handler: AwsBedrockHandler
@@ -46,6 +54,10 @@ describe("AwsBedrockHandler Error Handling", () => {
 		})
 	})
 
+	/**
+	 * Helper: create an Error with optional extra properties that
+	 * the production code inspects (status, name, $metadata, __type).
+	 */
 	const createMockError = (options: {
 		message?: string
 		name?: string
@@ -56,505 +68,481 @@ describe("AwsBedrockHandler Error Handling", () => {
 			requestId?: string
 			extendedRequestId?: string
 			cfId?: string
-			[key: string]: any // Allow additional properties
+			[key: string]: unknown
 		}
 	}): Error => {
 		const error = new Error(options.message || "Test error") as any
 		if (options.name) error.name = options.name
-		if (options.status) error.status = options.status
+		if (options.status !== undefined) error.status = options.status
 		if (options.__type) error.__type = options.__type
 		if (options.$metadata) error.$metadata = options.$metadata
 		return error
 	}
 
-	describe("Throttling Error Detection", () => {
-		it("should detect throttling from HTTP 429 status code", async () => {
+	// -----------------------------------------------------------------------
+	// Throttling Detection — completePrompt path
+	//
+	// Production flow: generateText throws → catch → isThrottlingError() is
+	// NOT called in completePrompt (only in createMessage), so it falls
+	// through to handleAiSdkError which wraps with "Bedrock: <msg>".
+	//
+	// For createMessage: streamText throws → catch → isThrottlingError()
+	// returns true → re-throws original error.
+	// -----------------------------------------------------------------------
+
+	describe("Throttling Error Detection (createMessage)", () => {
+		it("should re-throw throttling errors with status 429 for retry", async () => {
 			const throttleError = createMockError({
 				message: "Request failed",
 				status: 429,
 			})
 
-			mockSend.mockRejectedValueOnce(throttleError)
+			mockStreamText.mockImplementation(() => {
+				throw throttleError
+			})
 
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("throttled or rate limited")
-			} catch (error) {
-				expect(error.message).toContain("throttled or rate limited")
-			}
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
+
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
+				}
+			}).rejects.toThrow("Request failed")
 		})
 
-		it("should detect throttling from AWS SDK $metadata.httpStatusCode", async () => {
+		it("should re-throw throttling errors detected via $metadata.httpStatusCode", async () => {
 			const throttleError = createMockError({
 				message: "Request failed",
 				$metadata: { httpStatusCode: 429 },
 			})
 
-			mockSend.mockRejectedValueOnce(throttleError)
+			mockStreamText.mockImplementation(() => {
+				throw throttleError
+			})
 
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("throttled or rate limited")
-			} catch (error) {
-				expect(error.message).toContain("throttled or rate limited")
-			}
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
+
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
+				}
+			}).rejects.toThrow("Request failed")
 		})
 
-		it("should detect throttling from ThrottlingException name", async () => {
+		it("should re-throw ThrottlingException by name", async () => {
 			const throttleError = createMockError({
 				message: "Request failed",
 				name: "ThrottlingException",
 			})
 
-			mockSend.mockRejectedValueOnce(throttleError)
-
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("throttled or rate limited")
-			} catch (error) {
-				expect(error.message).toContain("throttled or rate limited")
-			}
-		})
-
-		it("should detect throttling from __type field", async () => {
-			const throttleError = createMockError({
-				message: "Request failed",
-				__type: "ThrottlingException",
+			mockStreamText.mockImplementation(() => {
+				throw throttleError
 			})
 
-			mockSend.mockRejectedValueOnce(throttleError)
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
 
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("throttled or rate limited")
-			} catch (error) {
-				expect(error.message).toContain("throttled or rate limited")
-			}
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
+				}
+			}).rejects.toThrow("Request failed")
 		})
 
-		it("should detect throttling from 'Bedrock is unable to process your request' message", async () => {
+		it("should re-throw 'Bedrock is unable to process your request' as throttling", async () => {
 			const throttleError = createMockError({
 				message: "Bedrock is unable to process your request",
 			})
 
-			mockSend.mockRejectedValueOnce(throttleError)
+			mockStreamText.mockImplementation(() => {
+				throw throttleError
+			})
+
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
 
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("throttled or rate limited")
-			} catch (error) {
-				expect(error.message).toMatch(/throttled or rate limited/)
-			}
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
+				}
+			}).rejects.toThrow("Bedrock is unable to process your request")
 		})
 
 		it("should detect throttling from various message patterns", async () => {
-			const throttlingMessages = [
-				"Request throttled",
-				"Rate limit exceeded",
-				"Too many requests",
-				"Service unavailable due to high demand",
-				"Server is overloaded",
-				"System is busy",
-				"Please wait and try again",
-			]
+			const throttlingMessages = ["Request throttled", "Rate limit exceeded", "Too many requests"]
 
 			for (const message of throttlingMessages) {
+				vi.clearAllMocks()
 				const throttleError = createMockError({ message })
-				mockSend.mockRejectedValueOnce(throttleError)
-
-				try {
-					await handler.completePrompt("test")
-					// Should not reach here as completePrompt should throw
-					throw new Error("Expected error to be thrown")
-				} catch (error) {
-					expect(error.message).toContain("throttled or rate limited")
-				}
+
+				mockStreamText.mockImplementation(() => {
+					throw throttleError
+				})
+
+				const localHandler = new AwsBedrockHandler({
+					apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
+					awsAccessKey: "test-access-key",
+					awsSecretKey: "test-secret-key",
+					awsRegion: "us-east-1",
+				})
+
+				const generator = localHandler.createMessage("system", [{ role: "user", content: "test" }])
+
+				// Throttling errors are re-thrown with original message for retry
+				await expect(async () => {
+					for await (const _chunk of generator) {
+						// should throw
+					}
+				}).rejects.toThrow(message)
 			}
 		})
 
-		it("should display appropriate error information for throttling errors", async () => {
-			const throttlingError = createMockError({
-				message: "Bedrock is unable to process your request",
-				name: "ThrottlingException",
+		it("should prioritize HTTP status 429 over message content for throttling", async () => {
+			const mixedError = createMockError({
+				message: "Some generic error message",
 				status: 429,
-				$metadata: {
-					httpStatusCode: 429,
-					requestId: "12345-abcde-67890",
-					extendedRequestId: "extended-12345",
-					cfId: "cf-12345",
-				},
 			})
 
-			mockSend.mockRejectedValueOnce(throttlingError)
+			mockStreamText.mockImplementation(() => {
+				throw mixedError
+			})
 
-			try {
-				await handler.completePrompt("test")
-				throw new Error("Expected error to be thrown")
-			} catch (error) {
-				// Should contain the main error message
-				expect(error.message).toContain("throttled or rate limited")
-			}
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
+
+			// Because status=429, it's throttling → re-throws original error
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
+				}
+			}).rejects.toThrow("Some generic error message")
 		})
-	})
 
-	describe("Service Quota Exceeded Detection", () => {
-		it("should detect service quota exceeded errors", async () => {
-			const quotaError = createMockError({
-				message: "Service quota exceeded for model requests",
+		it("should prioritize ThrottlingException name over message for throttling", async () => {
+			const specificError = createMockError({
+				message: "Some other error occurred",
+				name: "ThrottlingException",
 			})
 
-			mockSend.mockRejectedValueOnce(quotaError)
+			mockStreamText.mockImplementation(() => {
+				throw specificError
+			})
 
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("Service quota exceeded")
-			} catch (error) {
-				expect(error.message).toContain("Service quota exceeded")
-			}
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
+
+			// ThrottlingException → re-throws original for retry
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
+				}
+			}).rejects.toThrow("Some other error occurred")
 		})
 	})
 
-	describe("Model Not Ready Detection", () => {
-		it("should detect model not ready errors", async () => {
-			const modelError = createMockError({
-				message: "Model is not ready, please try again later",
+	// -----------------------------------------------------------------------
+	// Non-throttling errors (createMessage) are wrapped by handleAiSdkError
+	// -----------------------------------------------------------------------
+
+	describe("Non-throttling errors (createMessage)", () => {
+		it("should wrap non-throttling errors with provider name via handleAiSdkError", async () => {
+			const genericError = createMockError({
+				message: "Something completely unexpected happened",
 			})
 
-			mockSend.mockRejectedValueOnce(modelError)
+			mockStreamText.mockImplementation(() => {
+				throw genericError
+			})
 
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("Model is not ready")
-			} catch (error) {
-				expect(error.message).toContain("Model is not ready")
-			}
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
+
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
+				}
+			}).rejects.toThrow("Bedrock: Something completely unexpected happened")
 		})
-	})
 
-	describe("Internal Server Error Detection", () => {
-		it("should detect internal server errors", async () => {
-			const serverError = createMockError({
+		it("should preserve status code from non-throttling API errors", async () => {
+			const apiError = createMockError({
 				message: "Internal server error occurred",
+				status: 500,
 			})
 
-			mockSend.mockRejectedValueOnce(serverError)
+			mockStreamText.mockImplementation(() => {
+				throw apiError
+			})
+
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
 
 			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("internal server error")
-			} catch (error) {
-				expect(error.message).toContain("internal server error")
+				for await (const _chunk of generator) {
+					// should throw
+				}
+				throw new Error("Expected error to be thrown")
+			} catch (error: any) {
+				expect(error.message).toContain("Bedrock:")
+				expect(error.message).toContain("Internal server error occurred")
 			}
 		})
-	})
 
-	describe("Token Limit Detection", () => {
-		it("should detect enhanced token limit errors", async () => {
-			const tokenErrors = [
-				"Too many tokens in request",
-				"Token limit exceeded",
-				"Maximum context length reached",
-				"Context length exceeds limit",
-			]
-
-			for (const message of tokenErrors) {
-				const tokenError = createMockError({ message })
-				mockSend.mockRejectedValueOnce(tokenError)
-
-				try {
-					await handler.completePrompt("test")
-					// Should not reach here as completePrompt should throw
-					throw new Error("Expected error to be thrown")
-				} catch (error) {
-					// Either "Too many tokens" for token-specific errors or "throttled" for limit-related errors
-					expect(error.message).toMatch(/Too many tokens|throttled or rate limited/)
+		it("should handle validation errors (token limits) as non-throttling", async () => {
+			const tokenError = createMockError({
+				message: "Too many tokens in request",
+				name: "ValidationException",
+			})
+
+			mockStreamText.mockImplementation(() => {
+				throw tokenError
+			})
+
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
+
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
 				}
-			}
+			}).rejects.toThrow("Bedrock: Too many tokens in request")
 		})
 	})
 
+	// -----------------------------------------------------------------------
+	// Streaming context: errors mid-stream
+	// -----------------------------------------------------------------------
+
 	describe("Streaming Context Error Handling", () => {
-		it("should handle throttling errors in streaming context", async () => {
+		it("should re-throw throttling errors that occur mid-stream", async () => {
 			const throttleError = createMockError({
 				message: "Bedrock is unable to process your request",
 				status: 429,
 			})
 
-			const mockStream = {
-				[Symbol.asyncIterator]() {
-					return {
-						async next() {
-							throw throttleError
-						},
-					}
-				},
+			// Mock streamText to return an object whose fullStream throws mid-iteration
+			async function* failingStream() {
+				yield { type: "text-delta" as const, textDelta: "partial" }
+				throw throttleError
 			}
 
-			mockSend.mockResolvedValueOnce({ stream: mockStream })
+			mockStreamText.mockReturnValue({
+				fullStream: failingStream(),
+				usage: Promise.resolve({ inputTokens: 0, outputTokens: 0 }),
+				providerMetadata: Promise.resolve({}),
+			})
 
 			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
 
-			// For throttling errors, it should throw immediately without yielding chunks
-			// This allows the retry mechanism to catch and handle it
 			await expect(async () => {
-				for await (const chunk of generator) {
-					// Should not yield any chunks for throttling errors
+				for await (const _chunk of generator) {
+					// may yield partial text before throwing
 				}
 			}).rejects.toThrow("Bedrock is unable to process your request")
 		})
 
-		it("should yield error chunks for non-throttling errors in streaming context", async () => {
+		it("should wrap non-throttling errors that occur mid-stream via handleAiSdkError", async () => {
 			const genericError = createMockError({
 				message: "Some other error",
 				status: 500,
 			})
 
-			const mockStream = {
-				[Symbol.asyncIterator]() {
-					return {
-						async next() {
-							throw genericError
-						},
-					}
-				},
+			async function* failingStream() {
+				yield { type: "text-delta" as const, textDelta: "partial" }
+				throw genericError
 			}
 
-			mockSend.mockResolvedValueOnce({ stream: mockStream })
+			mockStreamText.mockReturnValue({
+				fullStream: failingStream(),
+				usage: Promise.resolve({ inputTokens: 0, outputTokens: 0 }),
+				providerMetadata: Promise.resolve({}),
+			})
 
 			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
 
-			const chunks: any[] = []
-			try {
-				for await (const chunk of generator) {
-					chunks.push(chunk)
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
 				}
-			} catch (error) {
-				// Expected to throw after yielding chunks
-			}
-
-			// Should have yielded error chunks before throwing for non-throttling errors
-			expect(
-				chunks.some((chunk) => chunk.type === "text" && chunk.text && chunk.text.includes("Some other error")),
-			).toBe(true)
+			}).rejects.toThrow("Bedrock: Some other error")
 		})
 	})
 
-	describe("Error Priority and Specificity", () => {
-		it("should prioritize HTTP status codes over message patterns", async () => {
-			// Error with both 429 status and generic message should be detected as throttling
-			const mixedError = createMockError({
-				message: "Some generic error message",
-				status: 429,
-			})
+	// -----------------------------------------------------------------------
+	// completePrompt errors — all go through handleAiSdkError (no throttle check)
+	// -----------------------------------------------------------------------
 
-			mockSend.mockRejectedValueOnce(mixedError)
+	describe("completePrompt error handling", () => {
+		it("should wrap errors with provider name for completePrompt", async () => {
+			mockGenerateText.mockRejectedValueOnce(new Error("Bedrock API failure"))
 
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("throttled or rate limited")
-			} catch (error) {
-				expect(error.message).toContain("throttled or rate limited")
-			}
+			await expect(handler.completePrompt("test")).rejects.toThrow("Bedrock: Bedrock API failure")
 		})
 
-		it("should prioritize AWS error types over message patterns", async () => {
-			// Error with ThrottlingException name but different message should still be throttling
-			const specificError = createMockError({
-				message: "Some other error occurred",
-				name: "ThrottlingException",
+		it("should wrap throttling-pattern errors with provider name for completePrompt", async () => {
+			const throttleError = createMockError({
+				message: "Bedrock is unable to process your request",
+				status: 429,
 			})
 
-			mockSend.mockRejectedValueOnce(specificError)
+			mockGenerateText.mockRejectedValueOnce(throttleError)
 
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("throttled or rate limited")
-			} catch (error) {
-				expect(error.message).toContain("throttled or rate limited")
-			}
+			// completePrompt does NOT have the throttle-rethrow path; it always uses handleAiSdkError
+			await expect(handler.completePrompt("test")).rejects.toThrow(
+				"Bedrock: Bedrock is unable to process your request",
+			)
 		})
-	})
 
-	describe("Unknown Error Fallback", () => {
-		it("should still show unknown error for truly unrecognized errors", async () => {
-			const unknownError = createMockError({
-				message: "Something completely unexpected happened",
-			})
+		it("should handle concurrent generateText failures", async () => {
+			const error = new Error("API failure")
+			mockGenerateText.mockRejectedValue(error)
 
-			mockSend.mockRejectedValueOnce(unknownError)
+			const promises = Array.from({ length: 5 }, () => handler.completePrompt("test"))
+			const results = await Promise.allSettled(promises)
 
-			try {
-				const result = await handler.completePrompt("test")
-				expect(result).toContain("Unknown Error")
-			} catch (error) {
-				expect(error.message).toContain("Unknown Error")
-			}
+			results.forEach((result) => {
+				expect(result.status).toBe("rejected")
+				if (result.status === "rejected") {
+					expect(result.reason.message).toContain("Bedrock:")
+				}
+			})
 		})
-	})
 
-	describe("Enhanced Error Throw for Retry System", () => {
-		it("should throw enhanced error messages for completePrompt to display in retry system", async () => {
-			const throttlingError = createMockError({
-				message: "Too many tokens, rate limited",
-				status: 429,
-				$metadata: {
-					httpStatusCode: 429,
-					requestId: "test-request-id-12345",
-				},
+		it("should preserve status code from API call errors in completePrompt", async () => {
+			const apiError = createMockError({
+				message: "Service unavailable",
+				status: 503,
 			})
-			mockSend.mockRejectedValueOnce(throttlingError)
+
+			mockGenerateText.mockRejectedValueOnce(apiError)
 
 			try {
 				await handler.completePrompt("test")
 				throw new Error("Expected error to be thrown")
-			} catch (error) {
-				// Should contain the verbose message template
-				expect(error.message).toContain("Request was throttled or rate limited")
-				// Should preserve original error properties
-				expect((error as any).status).toBe(429)
-				expect((error as any).$metadata.requestId).toBe("test-request-id-12345")
+			} catch (error: any) {
+				expect(error.message).toContain("Bedrock:")
+				expect(error.message).toContain("Service unavailable")
 			}
 		})
+	})
 
-		it("should throw enhanced error messages for createMessage streaming to display in retry system", async () => {
-			const tokenError = createMockError({
-				message: "Too many tokens in request",
-				name: "ValidationException",
-				$metadata: {
-					httpStatusCode: 400,
-					requestId: "token-error-id-67890",
-					extendedRequestId: "extended-12345",
-				},
-			})
-
-			const mockStream = {
-				[Symbol.asyncIterator]() {
-					return {
-						async next() {
-							throw tokenError
-						},
-					}
-				},
-			}
+	// -----------------------------------------------------------------------
+	// Telemetry
+	// -----------------------------------------------------------------------
+
+	describe("Error telemetry", () => {
+		it("should capture telemetry for createMessage errors", async () => {
+			mockStreamText.mockImplementation(() => {
+				throw new Error("Stream failure")
+			})
 
-			mockSend.mockResolvedValueOnce({ stream: mockStream })
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
 
-			try {
-				const stream = handler.createMessage("system", [{ role: "user", content: "test" }])
-				for await (const chunk of stream) {
-					// Should not reach here as it should throw an error
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
 				}
-				throw new Error("Expected error to be thrown")
-			} catch (error) {
-				// Should contain error codes (note: this will be caught by the non-throttling error path)
-				expect(error.message).toContain("Too many tokens")
-				// Should preserve original error properties
-				expect(error.name).toBe("ValidationException")
-				expect((error as any).$metadata.requestId).toBe("token-error-id-67890")
-			}
+			}).rejects.toThrow()
+
+			expect(mockCaptureException).toHaveBeenCalled()
 		})
-	})
 
-	describe("Edge Case Test Coverage", () => {
-		it("should handle concurrent throttling errors correctly", async () => {
-			const throttlingError = createMockError({
-				message: "Bedrock is unable to process your request",
+		it("should capture telemetry for completePrompt errors", async () => {
+			mockGenerateText.mockRejectedValueOnce(new Error("Generate failure"))
+
+			await expect(handler.completePrompt("test")).rejects.toThrow()
+
+			expect(mockCaptureException).toHaveBeenCalled()
+		})
+
+		it("should capture telemetry for throttling errors too", async () => {
+			const throttleError = createMockError({
+				message: "Rate limit exceeded",
 				status: 429,
 			})
 
-			// Setup multiple concurrent requests that will all fail with throttling
-			mockSend.mockRejectedValue(throttlingError)
-
-			// Execute multiple concurrent requests
-			const promises = Array.from({ length: 5 }, () => handler.completePrompt("test"))
+			mockStreamText.mockImplementation(() => {
+				throw throttleError
+			})
 
-			// All should throw with throttling error
-			const results = await Promise.allSettled(promises)
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
 
-			results.forEach((result) => {
-				expect(result.status).toBe("rejected")
-				if (result.status === "rejected") {
-					expect(result.reason.message).toContain("throttled or rate limited")
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
 				}
-			})
+			}).rejects.toThrow()
+
+			// Telemetry is captured even for throttling errors
+			expect(mockCaptureException).toHaveBeenCalled()
 		})
+	})
 
-		it("should handle mixed error scenarios with both throttling and other indicators", async () => {
-			// Error with both 429 status (throttling) and validation error message
-			const mixedError = createMockError({
-				message: "ValidationException: Your input is invalid, but also rate limited",
-				name: "ValidationException",
-				status: 429,
-				$metadata: {
-					httpStatusCode: 429,
-					requestId: "mixed-error-id",
-				},
-			})
+	// -----------------------------------------------------------------------
+	// Edge cases
+	// -----------------------------------------------------------------------
 
-			mockSend.mockRejectedValueOnce(mixedError)
+	describe("Edge Case Test Coverage", () => {
+		it("should handle non-Error objects thrown by generateText", async () => {
+			mockGenerateText.mockRejectedValueOnce("string error")
 
-			try {
-				await handler.completePrompt("test")
-			} catch (error) {
-				// Should be treated as throttling due to 429 status taking priority
-				expect(error.message).toContain("throttled or rate limited")
-				// Should still preserve metadata
-				expect((error as any).$metadata?.requestId).toBe("mixed-error-id")
-			}
+			await expect(handler.completePrompt("test")).rejects.toThrow("Bedrock: string error")
 		})
 
-		it("should handle rapid successive retries in streaming context", async () => {
-			const throttlingError = createMockError({
-				message: "ThrottlingException",
-				name: "ThrottlingException",
+		it("should handle non-Error objects thrown by streamText", async () => {
+			mockStreamText.mockImplementation(() => {
+				throw "string error"
 			})
 
-			// Mock stream that throws immediately
-			const mockStream = {
-				// eslint-disable-next-line require-yield
-				[Symbol.asyncIterator]: async function* () {
-					throw throttlingError
-				},
-			}
-
-			mockSend.mockResolvedValueOnce({ stream: mockStream })
-
-			const messages: Anthropic.Messages.MessageParam[] = [{ role: "user", content: "test" }]
+			const generator = handler.createMessage("system", [{ role: "user", content: "test" }])
 
-			try {
-				// Should throw immediately without yielding any chunks
-				const stream = handler.createMessage("", messages)
-				const chunks = []
-				for await (const chunk of stream) {
-					chunks.push(chunk)
+			// Non-Error values are not detected as throttling → handleAiSdkError path
+			await expect(async () => {
+				for await (const _chunk of generator) {
+					// should throw
 				}
-				// Should not reach here
-				expect(chunks).toHaveLength(0)
-			} catch (error) {
-				// Error should be thrown immediately for retry mechanism
-				// The error might be a TypeError if the stream iterator fails
-				expect(error).toBeDefined()
-				// The important thing is that it throws immediately without yielding chunks
-			}
+			}).rejects.toThrow("Bedrock: string error")
 		})
 
-		it("should validate error properties exist before accessing them", async () => {
-			// Error with unusual structure
-			const unusualError = {
-				message: "Error with unusual structure",
-				// Missing typical properties like name, status, etc.
-			}
-
-			mockSend.mockRejectedValueOnce(unusualError)
+		it("should handle errors with unusual structure gracefully", async () => {
+			const unusualError = { message: "Error with unusual structure" }
+			mockGenerateText.mockRejectedValueOnce(unusualError)
 
 			try {
 				await handler.completePrompt("test")
-			} catch (error) {
-				// Should handle gracefully without accessing undefined properties
-				expect(error.message).toContain("Unknown Error")
-				// Should not have undefined values in the error message
+				throw new Error("Expected error to be thrown")
+			} catch (error: any) {
+				// handleAiSdkError wraps with "Bedrock: ..."
+				expect(error.message).toContain("Bedrock:")
 				expect(error.message).not.toContain("undefined")
 			}
 		})
+
+		it("should handle concurrent throttling errors in streaming context", async () => {
+			const throttlingError = createMockError({
+				message: "Bedrock is unable to process your request",
+				status: 429,
+			})
+
+			mockStreamText.mockImplementation(() => {
+				throw throttlingError
+			})
+
+			// Execute multiple concurrent streaming requests
+			const promises = Array.from({ length: 3 }, async () => {
+				const localHandler = new AwsBedrockHandler({
+					apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
+					awsAccessKey: "test-access-key",
+					awsSecretKey: "test-secret-key",
+					awsRegion: "us-east-1",
+				})
+				const gen = localHandler.createMessage("system", [{ role: "user", content: "test" }])
+				for await (const _chunk of gen) {
+					// should throw
+				}
+			})
+
+			const results = await Promise.allSettled(promises)
+			results.forEach((result) => {
+				expect(result.status).toBe("rejected")
+				if (result.status === "rejected") {
+					// Throttling errors are re-thrown with original message
+					expect(result.reason.message).toBe("Bedrock is unable to process your request")
+				}
+			})
+		})
 	})
 })

src/api/providers/__tests__/bedrock-inference-profiles.spec.ts

@@ -4,18 +4,6 @@ import { AWS_INFERENCE_PROFILE_MAPPING } from "@roo-code/types"
 import { AwsBedrockHandler } from "../bedrock"
 import { ApiHandlerOptions } from "../../../shared/api"
 
-// Mock AWS SDK
-vitest.mock("@aws-sdk/client-bedrock-runtime", () => {
-	return {
-		BedrockRuntimeClient: vitest.fn().mockImplementation(() => ({
-			send: vitest.fn(),
-			config: { region: "us-east-1" },
-		})),
-		ConverseCommand: vitest.fn(),
-		ConverseStreamCommand: vitest.fn(),
-	}
-})
-
 describe("Amazon Bedrock Inference Profiles", () => {
 	// Helper function to create a handler with specific options
 	const createHandler = (options: Partial<ApiHandlerOptions> = {}) => {

src/api/providers/__tests__/bedrock-invokedModelId.spec.ts

@@ -1,350 +1,198 @@
 // npx vitest run src/api/providers/__tests__/bedrock-invokedModelId.spec.ts
 
-import { ApiHandlerOptions } from "../../../shared/api"
-
-import { AwsBedrockHandler, StreamEvent } from "../bedrock"
+// Mock TelemetryService before other imports
+vi.mock("@roo-code/telemetry", () => ({
+	TelemetryService: {
+		instance: {
+			captureException: vi.fn(),
+		},
+	},
+}))
 
-// Mock AWS SDK credential providers and Bedrock client
-vitest.mock("@aws-sdk/credential-providers", () => ({
-	fromIni: vitest.fn().mockReturnValue({
+// Mock AWS SDK credential providers
+vi.mock("@aws-sdk/credential-providers", () => ({
+	fromIni: vi.fn().mockReturnValue({
 		accessKeyId: "profile-access-key",
 		secretAccessKey: "profile-secret-key",
 	}),
 }))
 
-// Mock Smithy client
-vitest.mock("@smithy/smithy-client", () => ({
-	throwDefaultError: vitest.fn(),
+// Use vi.hoisted to define mock functions for AI SDK
+const { mockStreamText, mockGenerateText } = vi.hoisted(() => ({
+	mockStreamText: vi.fn(),
+	mockGenerateText: vi.fn(),
 }))
 
-// Create a mock send function that we can reference
-const mockSend = vitest.fn().mockImplementation(async () => {
+vi.mock("ai", async (importOriginal) => {
+	const actual = await importOriginal<typeof import("ai")>()
 	return {
-		$metadata: {
-			httpStatusCode: 200,
-			requestId: "mock-request-id",
-		},
-		stream: {
-			[Symbol.asyncIterator]: async function* () {
-				yield {
-					metadata: {
-						usage: {
-							inputTokens: 100,
-							outputTokens: 200,
-						},
-					},
-				}
-			},
-		},
+		...actual,
+		streamText: mockStreamText,
+		generateText: mockGenerateText,
 	}
 })
 
-// Mock AWS SDK modules
-vitest.mock("@aws-sdk/client-bedrock-runtime", () => {
-	return {
-		BedrockRuntimeClient: vitest.fn().mockImplementation(() => ({
-			send: mockSend,
-			config: { region: "us-east-1" },
-			middlewareStack: {
-				clone: () => ({ resolve: () => {} }),
-				use: () => {},
-			},
-		})),
-		ConverseStreamCommand: vitest.fn((params) => ({
-			...params,
-			input: params,
-			middlewareStack: {
-				clone: () => ({ resolve: () => {} }),
-				use: () => {},
-			},
-		})),
-		ConverseCommand: vitest.fn((params) => ({
-			...params,
-			input: params,
-			middlewareStack: {
-				clone: () => ({ resolve: () => {} }),
-				use: () => {},
-			},
-		})),
-	}
-})
+vi.mock("@ai-sdk/amazon-bedrock", () => ({
+	createAmazonBedrock: vi.fn(() => vi.fn(() => ({ modelId: "test", provider: "bedrock" }))),
+}))
+
+import { AwsBedrockHandler } from "../bedrock"
+import { bedrockModels } from "@roo-code/types"
 
 describe("AwsBedrockHandler with invokedModelId", () => {
 	beforeEach(() => {
-		vitest.clearAllMocks()
+		vi.clearAllMocks()
 	})
 
-	// Helper function to create a mock async iterable stream
-	function createMockStream(events: StreamEvent[]) {
-		return {
-			[Symbol.asyncIterator]: async function* () {
-				for (const event of events) {
-					yield event
-				}
-				// Always yield a metadata event at the end
-				yield {
-					metadata: {
-						usage: {
-							inputTokens: 100,
-							outputTokens: 200,
+	/**
+	 * Helper: set up mockStreamText to return a stream whose resolved
+	 * `providerMetadata` contains the given `invokedModelId` in the
+	 * `bedrock.trace.promptRouter` path.
+	 */
+	function setupMockStreamWithInvokedModelId(invokedModelId?: string) {
+		async function* mockFullStream() {
+			yield { type: "text-delta", text: "Hello" }
+			yield { type: "text-delta", text: ", world!" }
+		}
+
+		const providerMetadata = invokedModelId
+			? {
+					bedrock: {
+						trace: {
+							promptRouter: {
+								invokedModelId,
+							},
 						},
 					},
 				}
-			},
-		}
+			: {}
+
+		mockStreamText.mockReturnValue({
+			fullStream: mockFullStream(),
+			usage: Promise.resolve({ inputTokens: 100, outputTokens: 200 }),
+			providerMetadata: Promise.resolve(providerMetadata),
+		})
 	}
 
-	it("should update costModelConfig when invokedModelId is present in the stream", async () => {
-		// Create a handler with a custom ARN
-		const mockOptions: ApiHandlerOptions = {
+	it("should update costModelConfig when invokedModelId is present in providerMetadata", async () => {
+		// Create a handler with a custom ARN (prompt router)
+		const handler = new AwsBedrockHandler({
 			awsAccessKey: "test-access-key",
 			awsSecretKey: "test-secret-key",
 			awsRegion: "us-east-1",
 			awsCustomArn: "arn:aws:bedrock:us-west-2:123456789:default-prompt-router/anthropic.claude:1",
-		}
-
-		const handler = new AwsBedrockHandler(mockOptions)
+		})
 
-		// Verify that getModel returns the updated model info
+		// The default prompt router model should use sonnet pricing (inputPrice: 3)
 		const initialModel = handler.getModel()
-		//the default prompt router model has an input price of 3. After the stream is handled it should be updated to 8
 		expect(initialModel.info.inputPrice).toBe(3)
 
-		// Create a spy on the getModel
-		const getModelByIdSpy = vitest.spyOn(handler, "getModelById")
-
-		// Mock the stream to include an event with invokedModelId and usage metadata
-		mockSend.mockImplementationOnce(async () => {
-			return {
-				stream: createMockStream([
-					// First event with invokedModelId and usage metadata
-					{
-						trace: {
-							promptRouter: {
-								invokedModelId:
-									"arn:aws:bedrock:us-west-2:699475926481:inference-profile/us.anthropic.claude-3-opus-20240229-v1:0",
-								usage: {
-									inputTokens: 150,
-									outputTokens: 250,
-									cacheReadTokens: 0,
-									cacheWriteTokens: 0,
-								},
-							},
-						},
-					},
-					{
-						contentBlockStart: {
-							start: {
-								text: "Hello",
-							},
-							contentBlockIndex: 0,
-						},
-					},
-					{
-						contentBlockDelta: {
-							delta: {
-								text: ", world!",
-							},
-							contentBlockIndex: 0,
-						},
-					},
-				]),
-			}
-		})
+		// Spy on getModelById to verify the invoked model is looked up
+		const getModelByIdSpy = vi.spyOn(handler, "getModelById")
 
-		// Create a message generator
-		const messageGenerator = handler.createMessage("system prompt", [{ role: "user", content: "user message" }])
+		// Set up stream to include an invokedModelId pointing to Claude 3 Opus
+		setupMockStreamWithInvokedModelId(
+			"arn:aws:bedrock:us-west-2:699475926481:inference-profile/us.anthropic.claude-3-opus-20240229-v1:0",
+		)
 
-		// Collect all yielded events to verify usage events
+		// Consume the generator
 		const events = []
-		for await (const event of messageGenerator) {
+		for await (const event of handler.createMessage("system prompt", [{ role: "user", content: "user message" }])) {
 			events.push(event)
 		}
 
-		// Verify that getModelById was called with the id, not the full arn
+		// Verify that getModelById was called with the parsed model id and type
 		expect(getModelByIdSpy).toHaveBeenCalledWith("anthropic.claude-3-opus-20240229-v1:0", "inference-profile")
 
-		// Verify that getModel returns the updated model info
+		// After processing, getModel should return the invoked model's pricing (Opus: inputPrice 15)
 		const costModel = handler.getModel()
-		//expect(costModel.id).toBe("anthropic.claude-3-5-sonnet-20240620-v1:0")
 		expect(costModel.info.inputPrice).toBe(15)
 
-		// Verify that a usage event was emitted after updating the costModelConfig
-		const usageEvents = events.filter((event) => event.type === "usage")
+		// Verify that a usage event was emitted
+		const usageEvents = events.filter((e: any) => e.type === "usage")
 		expect(usageEvents.length).toBeGreaterThanOrEqual(1)
 
-		// The last usage event should have the token counts from the metadata
-		const lastUsageEvent = usageEvents[usageEvents.length - 1]
-		// Expect the usage event to include all token information
+		// The usage event should contain the token counts
+		const lastUsageEvent = usageEvents[usageEvents.length - 1] as any
 		expect(lastUsageEvent).toMatchObject({
 			type: "usage",
 			inputTokens: 100,
 			outputTokens: 200,
-			// Cache tokens may be present with default values
-			cacheReadTokens: expect.any(Number),
-			cacheWriteTokens: expect.any(Number),
 		})
 	})
 
 	it("should not update costModelConfig when invokedModelId is not present", async () => {
-		// Create a handler with default settings
-		const mockOptions: ApiHandlerOptions = {
+		const handler = new AwsBedrockHandler({
 			apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 			awsAccessKey: "test-access-key",
 			awsSecretKey: "test-secret-key",
 			awsRegion: "us-east-1",
-		}
-
-		const handler = new AwsBedrockHandler(mockOptions)
+		})
 
-		// Store the initial model configuration
 		const initialModelConfig = handler.getModel()
 		expect(initialModelConfig.id).toBe("anthropic.claude-3-5-sonnet-20241022-v2:0")
 
-		// Mock the stream without an invokedModelId event
-		mockSend.mockImplementationOnce(async () => {
-			return {
-				stream: createMockStream([
-					// Some content events but no invokedModelId
-					{
-						contentBlockStart: {
-							start: {
-								text: "Hello",
-							},
-							contentBlockIndex: 0,
-						},
-					},
-					{
-						contentBlockDelta: {
-							delta: {
-								text: ", world!",
-							},
-							contentBlockIndex: 0,
-						},
-					},
-				]),
-			}
-		})
-
-		// Create a message generator
-		const messageGenerator = handler.createMessage("system prompt", [{ role: "user", content: "user message" }])
+		// Set up stream WITHOUT an invokedModelId
+		setupMockStreamWithInvokedModelId(undefined)
 
 		// Consume the generator
-		for await (const _ of messageGenerator) {
-			// Just consume the messages
+		for await (const _ of handler.createMessage("system prompt", [{ role: "user", content: "user message" }])) {
+			// Just consume
 		}
 
-		// Verify that getModel returns the original model info (unchanged)
+		// Model should remain unchanged
 		const costModel = handler.getModel()
 		expect(costModel.id).toBe("anthropic.claude-3-5-sonnet-20241022-v2:0")
-		expect(costModel).toEqual(initialModelConfig)
+		expect(costModel.info.inputPrice).toBe(initialModelConfig.info.inputPrice)
 	})
 
 	it("should handle invalid invokedModelId format gracefully", async () => {
-		// Create a handler with default settings
-		const mockOptions: ApiHandlerOptions = {
+		const handler = new AwsBedrockHandler({
 			apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 			awsAccessKey: "test-access-key",
 			awsSecretKey: "test-secret-key",
 			awsRegion: "us-east-1",
-		}
-
-		const handler = new AwsBedrockHandler(mockOptions)
-
-		// Mock the stream with an invalid invokedModelId
-		mockSend.mockImplementationOnce(async () => {
-			return {
-				stream: createMockStream([
-					// Event with invalid invokedModelId format
-					{
-						trace: {
-							promptRouter: {
-								invokedModelId: "invalid-format-not-an-arn",
-							},
-						},
-					},
-					// Some content events
-					{
-						contentBlockStart: {
-							start: {
-								text: "Hello",
-							},
-							contentBlockIndex: 0,
-						},
-					},
-				]),
-			}
 		})
 
-		// Create a message generator
-		const messageGenerator = handler.createMessage("system prompt", [{ role: "user", content: "user message" }])
+		// Set up stream with an invalid (non-ARN) invokedModelId
+		setupMockStreamWithInvokedModelId("invalid-format-not-an-arn")
 
-		// Consume the generator
-		for await (const _ of messageGenerator) {
-			// Just consume the messages
+		// Consume the generator — should not throw
+		for await (const _ of handler.createMessage("system prompt", [{ role: "user", content: "user message" }])) {
+			// Just consume
 		}
 
-		// Verify that getModel returns the original model info
+		// Model should remain unchanged (the parseArn call should fail gracefully)
 		const costModel = handler.getModel()
 		expect(costModel.id).toBe("anthropic.claude-3-5-sonnet-20241022-v2:0")
 	})
 
-	it("should handle errors during invokedModelId processing", async () => {
-		// Create a handler with default settings
-		const mockOptions: ApiHandlerOptions = {
-			apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
+	it("should use the invoked model's pricing for totalCost calculation", async () => {
+		const handler = new AwsBedrockHandler({
 			awsAccessKey: "test-access-key",
 			awsSecretKey: "test-secret-key",
 			awsRegion: "us-east-1",
-		}
-
-		const handler = new AwsBedrockHandler(mockOptions)
-
-		// Mock the stream with a valid invokedModelId
-		mockSend.mockImplementationOnce(async () => {
-			return {
-				stream: createMockStream([
-					// Event with valid invokedModelId
-					{
-						trace: {
-							promptRouter: {
-								invokedModelId:
-									"arn:aws:bedrock:us-east-1:123456789:foundation-model/anthropic.claude-3-sonnet-20240229-v1:0",
-							},
-						},
-					},
-				]),
-			}
-		})
-
-		// Mock getModel to throw an error when called with the model name
-		vitest.spyOn(handler, "getModel").mockImplementation((modelName?: string) => {
-			if (modelName === "anthropic.claude-3-sonnet-20240229-v1:0") {
-				throw new Error("Test error during model lookup")
-			}
-
-			// Default return value for initial call
-			return {
-				id: "anthropic.claude-3-5-sonnet-20241022-v2:0",
-				info: {
-					maxTokens: 4096,
-					contextWindow: 128_000,
-					supportsPromptCache: false,
-					supportsImages: true,
-				},
-			}
+			awsCustomArn: "arn:aws:bedrock:us-west-2:123456789:default-prompt-router/anthropic.claude:1",
 		})
 
-		// Create a message generator
-		const messageGenerator = handler.createMessage("system prompt", [{ role: "user", content: "user message" }])
+		// Set up stream to include Opus as the invoked model
+		setupMockStreamWithInvokedModelId(
+			"arn:aws:bedrock:us-west-2:699475926481:foundation-model/anthropic.claude-3-opus-20240229-v1:0",
+		)
 
-		// Consume the generator
-		for await (const _ of messageGenerator) {
-			// Just consume the messages
+		const events = []
+		for await (const event of handler.createMessage("system prompt", [{ role: "user", content: "user message" }])) {
+			events.push(event)
 		}
 
-		// Verify that getModel returns the original model info
-		const costModel = handler.getModel()
-		expect(costModel.id).toBe("anthropic.claude-3-5-sonnet-20241022-v2:0")
+		const usageEvent = events.find((e: any) => e.type === "usage") as any
+		expect(usageEvent).toBeDefined()
+
+		// Calculate expected cost based on Opus pricing ($15 / 1M input, $75 / 1M output)
+		const opusInfo = bedrockModels["anthropic.claude-3-opus-20240229-v1:0"]
+		const expectedCost =
+			(100 * (opusInfo.inputPrice ?? 0)) / 1_000_000 + (200 * (opusInfo.outputPrice ?? 0)) / 1_000_000
+
+		expect(usageEvent.totalCost).toBeCloseTo(expectedCost, 10)
 	})
 })

src/api/providers/__tests__/bedrock-native-tools.spec.ts

@@ -1,3 +1,12 @@
+// Mock TelemetryService before other imports
+vi.mock("@roo-code/telemetry", () => ({
+	TelemetryService: {
+		instance: {
+			captureException: vi.fn(),
+		},
+	},
+}))
+
 // Mock AWS SDK credential providers
 vi.mock("@aws-sdk/credential-providers", () => {
 	const mockFromIni = vi.fn().mockReturnValue({
@@ -7,29 +16,28 @@ vi.mock("@aws-sdk/credential-providers", () => {
 	return { fromIni: mockFromIni }
 })
 
-// Mock BedrockRuntimeClient and ConverseStreamCommand
-const mockSend = vi.fn()
+// Use vi.hoisted to define mock functions for AI SDK
+const { mockStreamText, mockGenerateText } = vi.hoisted(() => ({
+	mockStreamText: vi.fn(),
+	mockGenerateText: vi.fn(),
+}))
 
-vi.mock("@aws-sdk/client-bedrock-runtime", () => {
+vi.mock("ai", async (importOriginal) => {
+	const actual = await importOriginal<typeof import("ai")>()
 	return {
-		BedrockRuntimeClient: vi.fn().mockImplementation(() => ({
-			send: mockSend,
-			config: { region: "us-east-1" },
-		})),
-		ConverseStreamCommand: vi.fn((params) => ({
-			...params,
-			input: params,
-		})),
-		ConverseCommand: vi.fn(),
+		...actual,
+		streamText: mockStreamText,
+		generateText: mockGenerateText,
 	}
 })
 
+vi.mock("@ai-sdk/amazon-bedrock", () => ({
+	createAmazonBedrock: vi.fn(() => vi.fn(() => ({ modelId: "test", provider: "bedrock" }))),
+}))
+
 import { AwsBedrockHandler } from "../bedrock"
-import { ConverseStreamCommand } from "@aws-sdk/client-bedrock-runtime"
 import type { ApiHandlerCreateMessageMetadata } from "../../index"
 
-const mockConverseStreamCommand = vi.mocked(ConverseStreamCommand)
-
 // Test tool definitions in OpenAI format
 const testTools = [
 	{
@@ -63,542 +71,529 @@ const testTools = [
 	},
 ]
 
-describe("AwsBedrockHandler Native Tool Calling", () => {
+/**
+ * Helper: set up mockStreamText to return a simple text-delta stream.
+ */
+function setupMockStreamText() {
+	async function* mockFullStream() {
+		yield { type: "text-delta", text: "Response text" }
+	}
+	mockStreamText.mockReturnValue({
+		fullStream: mockFullStream(),
+		usage: Promise.resolve({ inputTokens: 100, outputTokens: 50 }),
+		providerMetadata: Promise.resolve({}),
+	})
+}
+
+/**
+ * Helper: set up mockStreamText to return a stream with tool-call events.
+ */
+function setupMockStreamTextWithToolCall() {
+	async function* mockFullStream() {
+		yield {
+			type: "tool-input-start",
+			id: "tool-123",
+			toolName: "read_file",
+		}
+		yield {
+			type: "tool-input-delta",
+			id: "tool-123",
+			delta: '{"path": "/test.txt"}',
+		}
+		yield {
+			type: "tool-input-end",
+			id: "tool-123",
+		}
+	}
+	mockStreamText.mockReturnValue({
+		fullStream: mockFullStream(),
+		usage: Promise.resolve({ inputTokens: 100, outputTokens: 50 }),
+		providerMetadata: Promise.resolve({}),
+	})
+}
+
+describe("AwsBedrockHandler Native Tool Calling (AI SDK)", () => {
 	let handler: AwsBedrockHandler
 
 	beforeEach(() => {
 		vi.clearAllMocks()
 
-		// Create handler with a model that supports native tools
 		handler = new AwsBedrockHandler({
 			apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 			awsAccessKey: "test-access-key",
 			awsSecretKey: "test-secret-key",
 			awsRegion: "us-east-1",
 		})
-
-		// Mock the stream response
-		mockSend.mockResolvedValue({
-			stream: [],
-		})
 	})
 
-	describe("convertToolsForBedrock", () => {
-		it("should convert OpenAI tools to Bedrock format", () => {
-			// Access private method
-			const convertToolsForBedrock = (handler as any).convertToolsForBedrock.bind(handler)
-
-			const bedrockTools = convertToolsForBedrock(testTools)
-
-			expect(bedrockTools).toHaveLength(2)
-
-			// Check structure and key properties (normalizeToolSchema adds additionalProperties: false)
-			const tool = bedrockTools[0]
-			expect(tool.toolSpec.name).toBe("read_file")
-			expect(tool.toolSpec.description).toBe("Read a file from the filesystem")
-			expect(tool.toolSpec.inputSchema.json.type).toBe("object")
-			expect(tool.toolSpec.inputSchema.json.properties.path.type).toBe("string")
-			expect(tool.toolSpec.inputSchema.json.properties.path.description).toBe("The path to the file")
-			expect(tool.toolSpec.inputSchema.json.required).toEqual(["path"])
-			// normalizeToolSchema adds additionalProperties: false by default
-			expect(tool.toolSpec.inputSchema.json.additionalProperties).toBe(false)
+	describe("tools passed to streamText", () => {
+		it("should pass converted tools to streamText when tools are provided", async () => {
+			setupMockStreamText()
+
+			const metadata: ApiHandlerCreateMessageMetadata = {
+				taskId: "test-task",
+				tools: testTools,
+			}
+
+			const generator = handler.createMessage(
+				"You are a helpful assistant.",
+				[{ role: "user", content: "Read the file at /test.txt" }],
+				metadata,
+			)
+
+			// Drain the generator
+			for await (const _chunk of generator) {
+				/* consume */
+			}
+
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+
+			// tools should be defined and contain AI SDK tool objects keyed by name
+			expect(callArgs.tools).toBeDefined()
+			expect(callArgs.tools.read_file).toBeDefined()
+			expect(callArgs.tools.write_file).toBeDefined()
 		})
 
-		it("should transform type arrays to anyOf for JSON Schema 2020-12 compliance", () => {
-			const convertToolsForBedrock = (handler as any).convertToolsForBedrock.bind(handler)
-
-			// Tools with type: ["string", "null"] syntax (valid in draft-07 but not 2020-12)
-			const toolsWithNullableTypes = [
-				{
-					type: "function" as const,
-					function: {
-						name: "execute_command",
-						description: "Execute a command",
-						parameters: {
-							type: "object",
-							properties: {
-								command: { type: "string", description: "The command to execute" },
-								cwd: {
-									type: ["string", "null"],
-									description: "Working directory (optional)",
-								},
-							},
-							required: ["command", "cwd"],
-						},
-					},
-				},
-				{
-					type: "function" as const,
-					function: {
-						name: "read_file",
-						description: "Read files",
-						parameters: {
-							type: "object",
-							properties: {
-								path: { type: "string" },
-								indentation: {
-									type: ["object", "null"],
-									properties: {
-										anchor_line: {
-											type: ["integer", "null"],
-											description: "Optional anchor line",
-										},
-									},
-								},
-							},
-							required: ["path"],
-						},
-					},
-				},
-			]
+		it("should pass undefined tools when no tools are provided in metadata", async () => {
+			setupMockStreamText()
+
+			const metadata: ApiHandlerCreateMessageMetadata = {
+				taskId: "test-task",
+				// No tools
+			}
 
-			const bedrockTools = convertToolsForBedrock(toolsWithNullableTypes)
+			const generator = handler.createMessage(
+				"You are a helpful assistant.",
+				[{ role: "user", content: "Hello" }],
+				metadata,
+			)
 
-			expect(bedrockTools).toHaveLength(2)
+			for await (const _chunk of generator) {
+				/* consume */
+			}
 
-			// First tool: cwd should be transformed from type: ["string", "null"] to anyOf
-			const executeCommandSchema = bedrockTools[0].toolSpec.inputSchema.json as any
-			expect(executeCommandSchema.properties.cwd.anyOf).toEqual([{ type: "string" }, { type: "null" }])
-			expect(executeCommandSchema.properties.cwd.type).toBeUndefined()
-			expect(executeCommandSchema.properties.cwd.description).toBe("Working directory (optional)")
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
 
-			// Second tool: nested nullable object should be transformed from type: ["object", "null"] to anyOf
-			const readFileSchema = bedrockTools[1].toolSpec.inputSchema.json as any
-			const indentation = readFileSchema.properties.indentation
-			expect(indentation.anyOf).toBeDefined()
-			expect(indentation.type).toBeUndefined()
-			// Object-level schema properties are preserved at the root, not inside the anyOf object variant
-			expect(indentation.additionalProperties).toBe(false)
-			expect(indentation.properties.anchor_line.anyOf).toEqual([{ type: "integer" }, { type: "null" }])
+			// When no tools are provided, tools should be undefined
+			expect(callArgs.tools).toBeUndefined()
 		})
 
-		it("should filter non-function tools", () => {
-			const convertToolsForBedrock = (handler as any).convertToolsForBedrock.bind(handler)
+		it("should filter non-function tools before passing to streamText", async () => {
+			setupMockStreamText()
 
-			const mixedTools = [
+			const mixedTools: any[] = [
 				...testTools,
-				{ type: "other" as any, something: {} }, // Should be filtered out
+				{ type: "other", something: {} }, // Should be filtered out
 			]
 
-			const bedrockTools = convertToolsForBedrock(mixedTools)
+			const metadata: ApiHandlerCreateMessageMetadata = {
+				taskId: "test-task",
+				tools: mixedTools as any,
+			}
 
-			expect(bedrockTools).toHaveLength(2)
-		})
-	})
+			const generator = handler.createMessage(
+				"You are a helpful assistant.",
+				[{ role: "user", content: "Read a file" }],
+				metadata,
+			)
 
-	describe("convertToolChoiceForBedrock", () => {
-		it("should convert 'auto' to Bedrock auto format", () => {
-			const convertToolChoiceForBedrock = (handler as any).convertToolChoiceForBedrock.bind(handler)
+			for await (const _chunk of generator) {
+				/* consume */
+			}
 
-			const result = convertToolChoiceForBedrock("auto")
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
 
-			expect(result).toEqual({ auto: {} })
+			// Only function tools should be present (keyed by name)
+			expect(callArgs.tools).toBeDefined()
+			expect(Object.keys(callArgs.tools)).toHaveLength(2)
+			expect(callArgs.tools.read_file).toBeDefined()
+			expect(callArgs.tools.write_file).toBeDefined()
 		})
+	})
 
-		it("should convert 'required' to Bedrock any format", () => {
-			const convertToolChoiceForBedrock = (handler as any).convertToolChoiceForBedrock.bind(handler)
+	describe("toolChoice passed to streamText", () => {
+		it("should default toolChoice to undefined when tool_choice is not specified", async () => {
+			setupMockStreamText()
 
-			const result = convertToolChoiceForBedrock("required")
+			const metadata: ApiHandlerCreateMessageMetadata = {
+				taskId: "test-task",
+				tools: testTools,
+				// No tool_choice
+			}
 
-			expect(result).toEqual({ any: {} })
-		})
+			const generator = handler.createMessage(
+				"You are a helpful assistant.",
+				[{ role: "user", content: "Read the file" }],
+				metadata,
+			)
 
-		it("should return undefined for 'none'", () => {
-			const convertToolChoiceForBedrock = (handler as any).convertToolChoiceForBedrock.bind(handler)
+			for await (const _chunk of generator) {
+				/* consume */
+			}
 
-			const result = convertToolChoiceForBedrock("none")
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
 
-			expect(result).toBeUndefined()
+			// mapToolChoice(undefined) returns undefined
+			expect(callArgs.toolChoice).toBeUndefined()
 		})
 
-		it("should convert specific tool choice to Bedrock tool format", () => {
-			const convertToolChoiceForBedrock = (handler as any).convertToolChoiceForBedrock.bind(handler)
+		it("should pass 'auto' toolChoice when tool_choice is 'auto'", async () => {
+			setupMockStreamText()
 
-			const result = convertToolChoiceForBedrock({
-				type: "function",
-				function: { name: "read_file" },
-			})
+			const metadata: ApiHandlerCreateMessageMetadata = {
+				taskId: "test-task",
+				tools: testTools,
+				tool_choice: "auto",
+			}
 
-			expect(result).toEqual({
-				tool: {
-					name: "read_file",
-				},
-			})
-		})
+			const generator = handler.createMessage(
+				"You are a helpful assistant.",
+				[{ role: "user", content: "Read the file" }],
+				metadata,
+			)
 
-		it("should default to auto for undefined toolChoice", () => {
-			const convertToolChoiceForBedrock = (handler as any).convertToolChoiceForBedrock.bind(handler)
+			for await (const _chunk of generator) {
+				/* consume */
+			}
 
-			const result = convertToolChoiceForBedrock(undefined)
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
 
-			expect(result).toEqual({ auto: {} })
+			expect(callArgs.toolChoice).toBe("auto")
 		})
-	})
 
-	describe("createMessage with native tools", () => {
-		it("should include toolConfig when tools are provided", async () => {
-			const handlerWithNativeTools = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
-			})
+		it("should pass 'none' toolChoice when tool_choice is 'none'", async () => {
+			setupMockStreamText()
 
 			const metadata: ApiHandlerCreateMessageMetadata = {
 				taskId: "test-task",
 				tools: testTools,
+				tool_choice: "none",
 			}
 
-			const generator = handlerWithNativeTools.createMessage(
+			const generator = handler.createMessage(
 				"You are a helpful assistant.",
-				[{ role: "user", content: "Read the file at /test.txt" }],
+				[{ role: "user", content: "Read the file" }],
 				metadata,
 			)
 
-			await generator.next()
+			for await (const _chunk of generator) {
+				/* consume */
+			}
 
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
 
-			expect(commandArg.toolConfig).toBeDefined()
-			expect(commandArg.toolConfig.tools).toHaveLength(2)
-			expect(commandArg.toolConfig.tools[0].toolSpec.name).toBe("read_file")
-			expect(commandArg.toolConfig.toolChoice).toEqual({ auto: {} })
+			expect(callArgs.toolChoice).toBe("none")
 		})
 
-		it("should always include toolConfig (tools are always present after PR #10841)", async () => {
-			const handlerWithNativeTools = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
-			})
+		it("should pass 'required' toolChoice when tool_choice is 'required'", async () => {
+			setupMockStreamText()
 
 			const metadata: ApiHandlerCreateMessageMetadata = {
 				taskId: "test-task",
-				// Even without explicit tools, tools are always present (minimum 6 from ALWAYS_AVAILABLE_TOOLS)
+				tools: testTools,
+				tool_choice: "required",
 			}
 
-			const generator = handlerWithNativeTools.createMessage(
+			const generator = handler.createMessage(
 				"You are a helpful assistant.",
-				[{ role: "user", content: "Read the file at /test.txt" }],
+				[{ role: "user", content: "Read the file" }],
 				metadata,
 			)
 
-			await generator.next()
+			for await (const _chunk of generator) {
+				/* consume */
+			}
 
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
 
-			// Tools are now always present
-			expect(commandArg.toolConfig).toBeDefined()
-			expect(commandArg.toolConfig.tools).toBeDefined()
-			expect(commandArg.toolConfig.toolChoice).toEqual({ auto: {} })
+			expect(callArgs.toolChoice).toBe("required")
 		})
 
-		it("should include toolConfig with undefined toolChoice when tool_choice is none", async () => {
-			const handlerWithNativeTools = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
-			})
+		it("should pass specific tool choice when tool_choice names a function", async () => {
+			setupMockStreamText()
 
 			const metadata: ApiHandlerCreateMessageMetadata = {
 				taskId: "test-task",
 				tools: testTools,
-				tool_choice: "none", // Explicitly disable tool use
+				tool_choice: {
+					type: "function",
+					function: { name: "read_file" },
+				},
 			}
 
-			const generator = handlerWithNativeTools.createMessage(
+			const generator = handler.createMessage(
 				"You are a helpful assistant.",
-				[{ role: "user", content: "Read the file at /test.txt" }],
+				[{ role: "user", content: "Read the file" }],
 				metadata,
 			)
 
-			await generator.next()
+			for await (const _chunk of generator) {
+				/* consume */
+			}
 
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
 
-			// toolConfig is still provided but toolChoice is undefined for "none"
-			expect(commandArg.toolConfig).toBeDefined()
-			expect(commandArg.toolConfig.toolChoice).toBeUndefined()
+			expect(callArgs.toolChoice).toEqual({
+				type: "tool",
+				toolName: "read_file",
+			})
 		})
+	})
 
-		it("should include fine-grained tool streaming beta for Claude models with native tools", async () => {
-			const handlerWithNativeTools = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
-			})
+	describe("tool call streaming events", () => {
+		it("should yield tool_call_start, tool_call_delta, and tool_call_end for tool input stream", async () => {
+			setupMockStreamTextWithToolCall()
 
 			const metadata: ApiHandlerCreateMessageMetadata = {
 				taskId: "test-task",
 				tools: testTools,
 			}
 
-			const generator = handlerWithNativeTools.createMessage(
+			const generator = handler.createMessage(
 				"You are a helpful assistant.",
-				[{ role: "user", content: "Read the file at /test.txt" }],
+				[{ role: "user", content: "Read the file" }],
 				metadata,
 			)
 
-			await generator.next()
+			const results: any[] = []
+			for await (const chunk of generator) {
+				results.push(chunk)
+			}
+
+			// Should have tool_call_start chunk
+			const startChunks = results.filter((r) => r.type === "tool_call_start")
+			expect(startChunks).toHaveLength(1)
+			expect(startChunks[0]).toEqual({
+				type: "tool_call_start",
+				id: "tool-123",
+				name: "read_file",
+			})
 
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
+			// Should have tool_call_delta chunk
+			const deltaChunks = results.filter((r) => r.type === "tool_call_delta")
+			expect(deltaChunks).toHaveLength(1)
+			expect(deltaChunks[0]).toEqual({
+				type: "tool_call_delta",
+				id: "tool-123",
+				delta: '{"path": "/test.txt"}',
+			})
 
-			// Should include the fine-grained tool streaming beta
-			expect(commandArg.additionalModelRequestFields).toBeDefined()
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).toContain(
-				"fine-grained-tool-streaming-2025-05-14",
-			)
+			// Should have tool_call_end chunk
+			const endChunks = results.filter((r) => r.type === "tool_call_end")
+			expect(endChunks).toHaveLength(1)
+			expect(endChunks[0]).toEqual({
+				type: "tool_call_end",
+				id: "tool-123",
+			})
 		})
 
-		it("should always include fine-grained tool streaming beta for Claude models", async () => {
-			const handlerWithNativeTools = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
+		it("should handle mixed text and tool use content in stream", async () => {
+			async function* mockFullStream() {
+				yield { type: "text-delta", text: "Let me read that file for you." }
+				yield { type: "text-delta", text: " Here's what I found:" }
+				yield {
+					type: "tool-input-start",
+					id: "tool-789",
+					toolName: "read_file",
+				}
+				yield {
+					type: "tool-input-delta",
+					id: "tool-789",
+					delta: '{"path": "/example.txt"}',
+				}
+				yield {
+					type: "tool-input-end",
+					id: "tool-789",
+				}
+			}
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 150, outputTokens: 75 }),
+				providerMetadata: Promise.resolve({}),
 			})
 
 			const metadata: ApiHandlerCreateMessageMetadata = {
 				taskId: "test-task",
-				// No tools provided
+				tools: testTools,
 			}
 
-			const generator = handlerWithNativeTools.createMessage(
+			const generator = handler.createMessage(
 				"You are a helpful assistant.",
-				[{ role: "user", content: "Hello" }],
+				[{ role: "user", content: "Read the example file" }],
 				metadata,
 			)
 
-			await generator.next()
+			const results: any[] = []
+			for await (const chunk of generator) {
+				results.push(chunk)
+			}
 
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
+			// Should have text chunks
+			const textChunks = results.filter((r) => r.type === "text")
+			expect(textChunks).toHaveLength(2)
+			expect(textChunks[0].text).toBe("Let me read that file for you.")
+			expect(textChunks[1].text).toBe(" Here's what I found:")
 
-			// Should always include anthropic_beta with fine-grained-tool-streaming for Claude models
-			expect(commandArg.additionalModelRequestFields).toBeDefined()
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).toContain(
-				"fine-grained-tool-streaming-2025-05-14",
-			)
+			// Should have tool call start
+			const startChunks = results.filter((r) => r.type === "tool_call_start")
+			expect(startChunks).toHaveLength(1)
+			expect(startChunks[0].name).toBe("read_file")
+
+			// Should have tool call delta
+			const deltaChunks = results.filter((r) => r.type === "tool_call_delta")
+			expect(deltaChunks).toHaveLength(1)
+			expect(deltaChunks[0].delta).toBe('{"path": "/example.txt"}')
+
+			// Should have tool call end
+			const endChunks = results.filter((r) => r.type === "tool_call_end")
+			expect(endChunks).toHaveLength(1)
 		})
-	})
 
-	describe("tool call streaming events", () => {
-		it("should yield tool_call_partial for toolUse block start", async () => {
-			const handlerWithNativeTools = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
+		it("should handle multiple tool calls in a single stream", async () => {
+			async function* mockFullStream() {
+				yield {
+					type: "tool-input-start",
+					id: "tool-1",
+					toolName: "read_file",
+				}
+				yield {
+					type: "tool-input-delta",
+					id: "tool-1",
+					delta: '{"path": "/file1.txt"}',
+				}
+				yield {
+					type: "tool-input-end",
+					id: "tool-1",
+				}
+				yield {
+					type: "tool-input-start",
+					id: "tool-2",
+					toolName: "write_file",
+				}
+				yield {
+					type: "tool-input-delta",
+					id: "tool-2",
+					delta: '{"path": "/file2.txt", "content": "hello"}',
+				}
+				yield {
+					type: "tool-input-end",
+					id: "tool-2",
+				}
+			}
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 200, outputTokens: 100 }),
+				providerMetadata: Promise.resolve({}),
 			})
 
-			// Mock stream with tool use events
-			mockSend.mockResolvedValue({
-				stream: (async function* () {
-					yield {
-						contentBlockStart: {
-							contentBlockIndex: 0,
-							start: {
-								toolUse: {
-									toolUseId: "tool-123",
-									name: "read_file",
-								},
-							},
-						},
-					}
-					yield {
-						contentBlockDelta: {
-							contentBlockIndex: 0,
-							delta: {
-								toolUse: {
-									input: '{"path": "/test.txt"}',
-								},
-							},
-						},
-					}
-					yield {
-						metadata: {
-							usage: {
-								inputTokens: 100,
-								outputTokens: 50,
-							},
-						},
-					}
-				})(),
-			})
+			const metadata: ApiHandlerCreateMessageMetadata = {
+				taskId: "test-task",
+				tools: testTools,
+			}
 
-			const generator = handlerWithNativeTools.createMessage("You are a helpful assistant.", [
-				{ role: "user", content: "Read the file" },
-			])
+			const generator = handler.createMessage(
+				"You are a helpful assistant.",
+				[{ role: "user", content: "Read and write files" }],
+				metadata,
+			)
 
 			const results: any[] = []
 			for await (const chunk of generator) {
 				results.push(chunk)
 			}
 
-			// Should have tool_call_partial chunks
-			const toolCallChunks = results.filter((r) => r.type === "tool_call_partial")
-			expect(toolCallChunks).toHaveLength(2)
+			// Should have two tool_call_start chunks
+			const startChunks = results.filter((r) => r.type === "tool_call_start")
+			expect(startChunks).toHaveLength(2)
+			expect(startChunks[0].name).toBe("read_file")
+			expect(startChunks[1].name).toBe("write_file")
 
-			// First chunk should have id and name
-			expect(toolCallChunks[0]).toEqual({
-				type: "tool_call_partial",
-				index: 0,
-				id: "tool-123",
-				name: "read_file",
-				arguments: undefined,
-			})
+			// Should have two tool_call_delta chunks
+			const deltaChunks = results.filter((r) => r.type === "tool_call_delta")
+			expect(deltaChunks).toHaveLength(2)
 
-			// Second chunk should have arguments
-			expect(toolCallChunks[1]).toEqual({
-				type: "tool_call_partial",
-				index: 0,
-				id: undefined,
-				name: undefined,
-				arguments: '{"path": "/test.txt"}',
-			})
+			// Should have two tool_call_end chunks
+			const endChunks = results.filter((r) => r.type === "tool_call_end")
+			expect(endChunks).toHaveLength(2)
 		})
+	})
 
-		it("should yield tool_call_partial for contentBlock toolUse structure", async () => {
-			const handlerWithNativeTools = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
-			})
+	describe("tools schema normalization", () => {
+		it("should apply schema normalization (additionalProperties: false, strict: true) via convertToolsForOpenAI", async () => {
+			setupMockStreamText()
 
-			// Mock stream with alternative tool use structure
-			mockSend.mockResolvedValue({
-				stream: (async function* () {
-					yield {
-						contentBlockStart: {
-							contentBlockIndex: 0,
-							contentBlock: {
-								toolUse: {
-									toolUseId: "tool-456",
-									name: "write_file",
+			const metadata: ApiHandlerCreateMessageMetadata = {
+				taskId: "test-task",
+				tools: [
+					{
+						type: "function" as const,
+						function: {
+							name: "test_tool",
+							description: "A test tool",
+							parameters: {
+								type: "object",
+								properties: {
+									arg1: { type: "string" },
 								},
+								// Note: no "required" field and no "additionalProperties"
 							},
 						},
-					}
-					yield {
-						metadata: {
-							usage: {
-								inputTokens: 100,
-								outputTokens: 50,
-							},
-						},
-					}
-				})(),
-			})
+					},
+				],
+			}
 
-			const generator = handlerWithNativeTools.createMessage("You are a helpful assistant.", [
-				{ role: "user", content: "Write a file" },
-			])
+			const generator = handler.createMessage(
+				"You are a helpful assistant.",
+				[{ role: "user", content: "test" }],
+				metadata,
+			)
 
-			const results: any[] = []
-			for await (const chunk of generator) {
-				results.push(chunk)
+			for await (const _chunk of generator) {
+				/* consume */
 			}
 
-			// Should have tool_call_partial chunk
-			const toolCallChunks = results.filter((r) => r.type === "tool_call_partial")
-			expect(toolCallChunks).toHaveLength(1)
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
 
-			expect(toolCallChunks[0]).toEqual({
-				type: "tool_call_partial",
-				index: 0,
-				id: "tool-456",
-				name: "write_file",
-				arguments: undefined,
-			})
+			// The AI SDK tools should be keyed by tool name
+			expect(callArgs.tools).toBeDefined()
+			expect(callArgs.tools.test_tool).toBeDefined()
 		})
+	})
 
-		it("should handle mixed text and tool use content", async () => {
-			const handlerWithNativeTools = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
-			})
+	describe("usage metrics with tools", () => {
+		it("should yield usage chunk after tool call stream completes", async () => {
+			setupMockStreamTextWithToolCall()
 
-			// Mock stream with mixed content
-			mockSend.mockResolvedValue({
-				stream: (async function* () {
-					yield {
-						contentBlockStart: {
-							contentBlockIndex: 0,
-							start: {
-								text: "Let me read that file for you.",
-							},
-						},
-					}
-					yield {
-						contentBlockDelta: {
-							contentBlockIndex: 0,
-							delta: {
-								text: " Here's what I found:",
-							},
-						},
-					}
-					yield {
-						contentBlockStart: {
-							contentBlockIndex: 1,
-							start: {
-								toolUse: {
-									toolUseId: "tool-789",
-									name: "read_file",
-								},
-							},
-						},
-					}
-					yield {
-						contentBlockDelta: {
-							contentBlockIndex: 1,
-							delta: {
-								toolUse: {
-									input: '{"path": "/example.txt"}',
-								},
-							},
-						},
-					}
-					yield {
-						metadata: {
-							usage: {
-								inputTokens: 150,
-								outputTokens: 75,
-							},
-						},
-					}
-				})(),
-			})
+			const metadata: ApiHandlerCreateMessageMetadata = {
+				taskId: "test-task",
+				tools: testTools,
+			}
 
-			const generator = handlerWithNativeTools.createMessage("You are a helpful assistant.", [
-				{ role: "user", content: "Read the example file" },
-			])
+			const generator = handler.createMessage(
+				"You are a helpful assistant.",
+				[{ role: "user", content: "Read a file" }],
+				metadata,
+			)
 
 			const results: any[] = []
 			for await (const chunk of generator) {
 				results.push(chunk)
 			}
 
-			// Should have text chunks
-			const textChunks = results.filter((r) => r.type === "text")
-			expect(textChunks).toHaveLength(2)
-			expect(textChunks[0].text).toBe("Let me read that file for you.")
-			expect(textChunks[1].text).toBe(" Here's what I found:")
-
-			// Should have tool call chunks
-			const toolCallChunks = results.filter((r) => r.type === "tool_call_partial")
-			expect(toolCallChunks).toHaveLength(2)
-			expect(toolCallChunks[0].name).toBe("read_file")
-			expect(toolCallChunks[1].arguments).toBe('{"path": "/example.txt"}')
+			// Should have a usage chunk at the end
+			const usageChunks = results.filter((r) => r.type === "usage")
+			expect(usageChunks).toHaveLength(1)
+			expect(usageChunks[0].inputTokens).toBe(100)
+			expect(usageChunks[0].outputTokens).toBe(50)
 		})
 	})
 })

src/api/providers/__tests__/bedrock-reasoning.spec.ts

@@ -1,37 +1,48 @@
-// npx vitest api/providers/__tests__/bedrock-reasoning.test.ts
+// npx vitest run api/providers/__tests__/bedrock-reasoning.spec.ts
+
+// Use vi.hoisted to define mock functions for AI SDK
+const { mockStreamText, mockGenerateText, mockCreateAmazonBedrock } = vi.hoisted(() => ({
+	mockStreamText: vi.fn(),
+	mockGenerateText: vi.fn(),
+	mockCreateAmazonBedrock: vi.fn(() => vi.fn(() => ({ modelId: "test", provider: "bedrock" }))),
+}))
+
+vi.mock("ai", async (importOriginal) => {
+	const actual = await importOriginal<typeof import("ai")>()
+	return {
+		...actual,
+		streamText: mockStreamText,
+		generateText: mockGenerateText,
+	}
+})
+
+vi.mock("@ai-sdk/amazon-bedrock", () => ({
+	createAmazonBedrock: mockCreateAmazonBedrock,
+}))
+
+// Mock AWS SDK credential providers
+vi.mock("@aws-sdk/credential-providers", () => ({
+	fromIni: vi.fn().mockReturnValue(async () => ({
+		accessKeyId: "profile-access-key",
+		secretAccessKey: "profile-secret-key",
+	})),
+}))
+
+vi.mock("../../../utils/logging", () => ({
+	logger: {
+		info: vi.fn(),
+		error: vi.fn(),
+		warn: vi.fn(),
+		debug: vi.fn(),
+	},
+}))
 
 import { AwsBedrockHandler } from "../bedrock"
-import { BedrockRuntimeClient, ConverseStreamCommand } from "@aws-sdk/client-bedrock-runtime"
 import { logger } from "../../../utils/logging"
 
-// Mock the AWS SDK
-vi.mock("@aws-sdk/client-bedrock-runtime")
-vi.mock("../../../utils/logging")
-
-// Store the command payload for verification
-let capturedPayload: any = null
-
 describe("AwsBedrockHandler - Extended Thinking", () => {
-	let handler: AwsBedrockHandler
-	let mockSend: ReturnType<typeof vi.fn>
-
 	beforeEach(() => {
-		capturedPayload = null
-		mockSend = vi.fn()
-
-		// Mock ConverseStreamCommand to capture the payload
-		;(ConverseStreamCommand as unknown as ReturnType<typeof vi.fn>).mockImplementation((payload) => {
-			capturedPayload = payload
-			return {
-				input: payload,
-			}
-		})
-		;(BedrockRuntimeClient as unknown as ReturnType<typeof vi.fn>).mockImplementation(() => ({
-			send: mockSend,
-			config: { region: "us-east-1" },
-		}))
-		;(logger.info as unknown as ReturnType<typeof vi.fn>).mockImplementation(() => {})
-		;(logger.error as unknown as ReturnType<typeof vi.fn>).mockImplementation(() => {})
+		vi.clearAllMocks()
 	})
 
 	afterEach(() => {
@@ -39,8 +50,8 @@ describe("AwsBedrockHandler - Extended Thinking", () => {
 	})
 
 	describe("Extended Thinking Support", () => {
-		it("should include thinking parameter for Claude Sonnet 4 when reasoning is enabled", async () => {
-			handler = new AwsBedrockHandler({
+		it("should include reasoningConfig in providerOptions when reasoning is enabled", async () => {
+			const handler = new AwsBedrockHandler({
 				apiProvider: "bedrock",
 				apiModelId: "anthropic.claude-sonnet-4-20250514-v1:0",
 				awsRegion: "us-east-1",
@@ -49,35 +60,17 @@ describe("AwsBedrockHandler - Extended Thinking", () => {
 				modelMaxThinkingTokens: 4096,
 			})
 
-			// Mock the stream response
-			mockSend.mockResolvedValue({
-				stream: (async function* () {
-					yield {
-						messageStart: { role: "assistant" },
-					}
-					yield {
-						contentBlockStart: {
-							content_block: { type: "thinking", thinking: "Let me think..." },
-							contentBlockIndex: 0,
-						},
-					}
-					yield {
-						contentBlockDelta: {
-							delta: { type: "thinking_delta", thinking: " about this problem." },
-						},
-					}
-					yield {
-						contentBlockStart: {
-							start: { text: "Here's the answer:" },
-							contentBlockIndex: 1,
-						},
-					}
-					yield {
-						metadata: {
-							usage: { inputTokens: 100, outputTokens: 50 },
-						},
-					}
-				})(),
+			// Mock stream with reasoning content
+			async function* mockFullStream() {
+				yield { type: "reasoning", text: "Let me think..." }
+				yield { type: "reasoning", text: " about this problem." }
+				yield { type: "text-delta", text: "Here's the answer:" }
+			}
+
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 100, outputTokens: 50 }),
+				providerMetadata: Promise.resolve({}),
 			})
 
 			const messages = [{ role: "user" as const, content: "Test message" }]
@@ -88,13 +81,14 @@ describe("AwsBedrockHandler - Extended Thinking", () => {
 				chunks.push(chunk)
 			}
 
-			// Verify the command was called with the correct payload
-			expect(mockSend).toHaveBeenCalledTimes(1)
-			expect(capturedPayload).toBeDefined()
-			expect(capturedPayload.additionalModelRequestFields).toBeDefined()
-			expect(capturedPayload.additionalModelRequestFields.thinking).toEqual({
+			// Verify streamText was called with providerOptions containing reasoningConfig
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+			expect(callArgs.providerOptions).toBeDefined()
+			expect(callArgs.providerOptions.bedrock).toBeDefined()
+			expect(callArgs.providerOptions.bedrock.reasoningConfig).toEqual({
 				type: "enabled",
-				budget_tokens: 4096, // Uses the full modelMaxThinkingTokens value
+				budgetTokens: 4096,
 			})
 
 			// Verify reasoning chunks were yielded
@@ -102,157 +96,145 @@ describe("AwsBedrockHandler - Extended Thinking", () => {
 			expect(reasoningChunks).toHaveLength(2)
 			expect(reasoningChunks[0].text).toBe("Let me think...")
 			expect(reasoningChunks[1].text).toBe(" about this problem.")
-
-			// Verify that topP is NOT present when thinking is enabled
-			expect(capturedPayload.inferenceConfig).not.toHaveProperty("topP")
 		})
 
-		it("should pass thinking parameters from metadata", async () => {
-			handler = new AwsBedrockHandler({
+		it("should not include reasoningConfig when reasoning is disabled", async () => {
+			const handler = new AwsBedrockHandler({
 				apiProvider: "bedrock",
 				apiModelId: "anthropic.claude-3-7-sonnet-20250219-v1:0",
 				awsRegion: "us-east-1",
+				// Note: no enableReasoningEffort = true, so thinking is disabled
 			})
 
-			mockSend.mockResolvedValue({
-				stream: (async function* () {
-					yield { messageStart: { role: "assistant" } }
-					yield { metadata: { usage: { inputTokens: 100, outputTokens: 50 } } }
-				})(),
+			async function* mockFullStream() {
+				yield { type: "text-delta", text: "Hello world" }
+			}
+
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 100, outputTokens: 50 }),
+				providerMetadata: Promise.resolve({}),
 			})
 
 			const messages = [{ role: "user" as const, content: "Test message" }]
-			const metadata = {
-				taskId: "test-task",
-				thinking: {
-					enabled: true,
-					maxTokens: 16384,
-					maxThinkingTokens: 8192,
-				},
-			}
+			const stream = handler.createMessage("System prompt", messages)
 
-			const stream = handler.createMessage("System prompt", messages, metadata)
 			const chunks = []
 			for await (const chunk of stream) {
 				chunks.push(chunk)
 			}
 
-			// Verify the thinking parameter was passed correctly
-			expect(mockSend).toHaveBeenCalledTimes(1)
-			expect(capturedPayload).toBeDefined()
-			expect(capturedPayload.additionalModelRequestFields).toBeDefined()
-			expect(capturedPayload.additionalModelRequestFields.thinking).toEqual({
-				type: "enabled",
-				budget_tokens: 8192,
-			})
-
-			// Verify that topP is NOT present when thinking is enabled via metadata
-			expect(capturedPayload.inferenceConfig).not.toHaveProperty("topP")
+			// Verify streamText was called — providerOptions should not contain reasoningConfig
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+			const bedrockOpts = callArgs.providerOptions?.bedrock
+			expect(bedrockOpts?.reasoningConfig).toBeUndefined()
 		})
 
-		it("should log when extended thinking is enabled", async () => {
-			handler = new AwsBedrockHandler({
+		it("should capture thinking signature from stream providerMetadata", async () => {
+			const handler = new AwsBedrockHandler({
 				apiProvider: "bedrock",
-				apiModelId: "anthropic.claude-opus-4-20250514-v1:0",
+				apiModelId: "anthropic.claude-sonnet-4-20250514-v1:0",
 				awsRegion: "us-east-1",
 				enableReasoningEffort: true,
-				modelMaxThinkingTokens: 5000,
+				modelMaxThinkingTokens: 4096,
 			})
 
-			mockSend.mockResolvedValue({
-				stream: (async function* () {
-					yield { messageStart: { role: "assistant" } }
-				})(),
+			const testSignature = "test-thinking-signature-abc123"
+
+			// Mock stream with reasoning content that includes a signature in providerMetadata
+			async function* mockFullStream() {
+				yield { type: "reasoning", text: "Let me think..." }
+				// The SDK emits signature as a reasoning-delta with providerMetadata.bedrock.signature
+				yield {
+					type: "reasoning",
+					text: "",
+					providerMetadata: { bedrock: { signature: testSignature } },
+				}
+				yield { type: "text-delta", text: "Answer" }
+			}
+
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 100, outputTokens: 50 }),
+				providerMetadata: Promise.resolve({}),
 			})
 
-			const messages = [{ role: "user" as const, content: "Test" }]
+			const messages = [{ role: "user" as const, content: "Test message" }]
 			const stream = handler.createMessage("System prompt", messages)
 
-			for await (const chunk of stream) {
+			for await (const _chunk of stream) {
 				// consume stream
 			}
 
-			// Verify logging
-			expect(logger.info).toHaveBeenCalledWith(
-				expect.stringContaining("Extended thinking enabled"),
-				expect.objectContaining({
-					ctx: "bedrock",
-					modelId: "anthropic.claude-opus-4-20250514-v1:0",
-				}),
-			)
+			// Verify thinking signature was captured
+			expect(handler.getThoughtSignature()).toBe(testSignature)
 		})
 
-		it("should not include topP when thinking is disabled (global removal)", async () => {
-			handler = new AwsBedrockHandler({
+		it("should capture redacted thinking blocks from stream providerMetadata", async () => {
+			const handler = new AwsBedrockHandler({
 				apiProvider: "bedrock",
-				apiModelId: "anthropic.claude-3-7-sonnet-20250219-v1:0",
+				apiModelId: "anthropic.claude-sonnet-4-20250514-v1:0",
 				awsRegion: "us-east-1",
-				// Note: no enableReasoningEffort = true, so thinking is disabled
+				enableReasoningEffort: true,
+				modelMaxThinkingTokens: 4096,
 			})
 
-			mockSend.mockResolvedValue({
-				stream: (async function* () {
-					yield { messageStart: { role: "assistant" } }
-					yield {
-						contentBlockStart: {
-							start: { text: "Hello" },
-							contentBlockIndex: 0,
-						},
-					}
-					yield {
-						contentBlockDelta: {
-							delta: { text: " world" },
-						},
-					}
-					yield { metadata: { usage: { inputTokens: 100, outputTokens: 50 } } }
-				})(),
+			const redactedData = "base64-encoded-redacted-data"
+
+			// Mock stream with redacted reasoning content
+			async function* mockFullStream() {
+				yield { type: "reasoning", text: "Some thinking..." }
+				yield {
+					type: "reasoning",
+					text: "",
+					providerMetadata: { bedrock: { redactedData } },
+				}
+				yield { type: "text-delta", text: "Answer" }
+			}
+
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 100, outputTokens: 50 }),
+				providerMetadata: Promise.resolve({}),
 			})
 
 			const messages = [{ role: "user" as const, content: "Test message" }]
 			const stream = handler.createMessage("System prompt", messages)
 
-			const chunks = []
-			for await (const chunk of stream) {
-				chunks.push(chunk)
+			for await (const _chunk of stream) {
+				// consume stream
 			}
 
-			// Verify that topP is NOT present for any model (removed globally)
-			expect(mockSend).toHaveBeenCalledTimes(1)
-			expect(capturedPayload).toBeDefined()
-			expect(capturedPayload.inferenceConfig).not.toHaveProperty("topP")
-
-			// Verify that additionalModelRequestFields contains fine-grained-tool-streaming for Claude models
-			expect(capturedPayload.additionalModelRequestFields).toBeDefined()
-			expect(capturedPayload.additionalModelRequestFields.anthropic_beta).toContain(
-				"fine-grained-tool-streaming-2025-05-14",
-			)
+			// Verify redacted thinking blocks were captured
+			const redactedBlocks = handler.getRedactedThinkingBlocks()
+			expect(redactedBlocks).toBeDefined()
+			expect(redactedBlocks).toHaveLength(1)
+			expect(redactedBlocks![0]).toEqual({
+				type: "redacted_thinking",
+				data: redactedData,
+			})
 		})
 
 		it("should enable reasoning when enableReasoningEffort is true in settings", async () => {
-			handler = new AwsBedrockHandler({
+			const handler = new AwsBedrockHandler({
 				apiProvider: "bedrock",
 				apiModelId: "anthropic.claude-sonnet-4-20250514-v1:0",
 				awsRegion: "us-east-1",
-				enableReasoningEffort: true, // This should trigger reasoning
+				enableReasoningEffort: true,
 				modelMaxThinkingTokens: 4096,
 			})
 
-			mockSend.mockResolvedValue({
-				stream: (async function* () {
-					yield { messageStart: { role: "assistant" } }
-					yield {
-						contentBlockStart: {
-							content_block: { type: "thinking", thinking: "Let me think..." },
-							contentBlockIndex: 0,
-						},
-					}
-					yield {
-						contentBlockDelta: {
-							delta: { type: "thinking_delta", thinking: " about this problem." },
-						},
-					}
-					yield { metadata: { usage: { inputTokens: 100, outputTokens: 50 } } }
-				})(),
+			async function* mockFullStream() {
+				yield { type: "reasoning", text: "Let me think..." }
+				yield { type: "reasoning", text: " about this problem." }
+				yield { type: "text-delta", text: "Test response" }
+			}
+
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 100, outputTokens: 50 }),
+				providerMetadata: Promise.resolve({}),
 			})
 
 			const messages = [{ role: "user" as const, content: "Test message" }]
@@ -264,17 +246,13 @@ describe("AwsBedrockHandler - Extended Thinking", () => {
 			}
 
 			// Verify thinking was enabled via settings
-			expect(mockSend).toHaveBeenCalledTimes(1)
-			expect(capturedPayload).toBeDefined()
-			expect(capturedPayload.additionalModelRequestFields).toBeDefined()
-			expect(capturedPayload.additionalModelRequestFields.thinking).toEqual({
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+			expect(callArgs.providerOptions?.bedrock?.reasoningConfig).toEqual({
 				type: "enabled",
-				budget_tokens: 4096,
+				budgetTokens: 4096,
 			})
 
-			// Verify that topP is NOT present when thinking is enabled via settings
-			expect(capturedPayload.inferenceConfig).not.toHaveProperty("topP")
-
 			// Verify reasoning chunks were yielded
 			const reasoningChunks = chunks.filter((c) => c.type === "reasoning")
 			expect(reasoningChunks).toHaveLength(2)
@@ -282,8 +260,8 @@ describe("AwsBedrockHandler - Extended Thinking", () => {
 			expect(reasoningChunks[1].text).toBe(" about this problem.")
 		})
 
-		it("should support API key authentication", async () => {
-			handler = new AwsBedrockHandler({
+		it("should support API key authentication via createAmazonBedrock", () => {
+			new AwsBedrockHandler({
 				apiProvider: "bedrock",
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsRegion: "us-east-1",
@@ -291,41 +269,13 @@ describe("AwsBedrockHandler - Extended Thinking", () => {
 				awsApiKey: "test-api-key-token",
 			})
 
-			mockSend.mockResolvedValue({
-				stream: (async function* () {
-					yield { messageStart: { role: "assistant" } }
-					yield {
-						contentBlockStart: {
-							start: { text: "Hello from API key auth" },
-							contentBlockIndex: 0,
-						},
-					}
-					yield { metadata: { usage: { inputTokens: 100, outputTokens: 50 } } }
-				})(),
-			})
-
-			const messages = [{ role: "user" as const, content: "Test message" }]
-			const stream = handler.createMessage("System prompt", messages)
-
-			const chunks = []
-			for await (const chunk of stream) {
-				chunks.push(chunk)
-			}
-
-			// Verify the client was created with API key token
-			expect(BedrockRuntimeClient).toHaveBeenCalledWith(
+			// Verify the provider was created with API key
+			expect(mockCreateAmazonBedrock).toHaveBeenCalledWith(
 				expect.objectContaining({
 					region: "us-east-1",
-					token: { token: "test-api-key-token" },
-					authSchemePreference: ["httpBearerAuth"],
+					apiKey: "test-api-key-token",
 				}),
 			)
-
-			// Verify the stream worked correctly
-			expect(mockSend).toHaveBeenCalledTimes(1)
-			const textChunks = chunks.filter((c) => c.type === "text")
-			expect(textChunks).toHaveLength(1)
-			expect(textChunks[0].text).toBe("Hello from API key auth")
 		})
 	})
 })

src/api/providers/__tests__/bedrock-vpc-endpoint.spec.ts

@@ -7,38 +7,40 @@ vi.mock("@aws-sdk/credential-providers", () => {
 	return { fromIni: mockFromIni }
 })
 
-// Mock BedrockRuntimeClient and ConverseStreamCommand
-vi.mock("@aws-sdk/client-bedrock-runtime", () => {
-	const mockSend = vi.fn().mockResolvedValue({
-		stream: [],
-	})
-	const mockBedrockRuntimeClient = vi.fn().mockImplementation(() => ({
-		send: mockSend,
-	}))
-
+// Use vi.hoisted to define mock functions for AI SDK
+const { mockStreamText, mockGenerateText } = vi.hoisted(() => ({
+	mockStreamText: vi.fn(),
+	mockGenerateText: vi.fn(),
+}))
+
+vi.mock("ai", async (importOriginal) => {
+	const actual = await importOriginal<typeof import("ai")>()
 	return {
-		BedrockRuntimeClient: mockBedrockRuntimeClient,
-		ConverseStreamCommand: vi.fn(),
-		ConverseCommand: vi.fn(),
+		...actual,
+		streamText: mockStreamText,
+		generateText: mockGenerateText,
 	}
 })
 
-import { AwsBedrockHandler } from "../bedrock"
-import { BedrockRuntimeClient } from "@aws-sdk/client-bedrock-runtime"
+// Mock createAmazonBedrock so we can inspect how it was called
+const { mockCreateAmazonBedrock } = vi.hoisted(() => ({
+	mockCreateAmazonBedrock: vi.fn(() => vi.fn(() => ({ modelId: "test", provider: "bedrock" }))),
+}))
 
-// Get access to the mocked functions
-const mockBedrockRuntimeClient = vi.mocked(BedrockRuntimeClient)
+vi.mock("@ai-sdk/amazon-bedrock", () => ({
+	createAmazonBedrock: mockCreateAmazonBedrock,
+}))
+
+import { AwsBedrockHandler } from "../bedrock"
 
 describe("Amazon Bedrock VPC Endpoint Functionality", () => {
 	beforeEach(() => {
-		// Clear all mocks before each test
 		vi.clearAllMocks()
 	})
 
 	// Test Scenario 1: Input Validation Test
 	describe("VPC Endpoint URL Validation", () => {
-		it("should configure client with endpoint URL when both URL and enabled flag are provided", () => {
-			// Create handler with endpoint URL and enabled flag
+		it("should configure provider with baseURL when both URL and enabled flag are provided", () => {
 			new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test-access-key",
@@ -48,17 +50,15 @@ describe("Amazon Bedrock VPC Endpoint Functionality", () => {
 				awsBedrockEndpointEnabled: true,
 			})
 
-			// Verify the client was created with the correct endpoint
-			expect(mockBedrockRuntimeClient).toHaveBeenCalledWith(
+			expect(mockCreateAmazonBedrock).toHaveBeenCalledWith(
 				expect.objectContaining({
 					region: "us-east-1",
-					endpoint: "https://bedrock-vpc.example.com",
+					baseURL: "https://bedrock-vpc.example.com",
 				}),
 			)
 		})
 
-		it("should not configure client with endpoint URL when URL is provided but enabled flag is false", () => {
-			// Create handler with endpoint URL but disabled flag
+		it("should not configure provider with baseURL when URL is provided but enabled flag is false", () => {
 			new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test-access-key",
@@ -68,23 +68,23 @@ describe("Amazon Bedrock VPC Endpoint Functionality", () => {
 				awsBedrockEndpointEnabled: false,
 			})
 
-			// Verify the client was created without the endpoint
-			expect(mockBedrockRuntimeClient).toHaveBeenCalledWith(
+			expect(mockCreateAmazonBedrock).toHaveBeenCalledWith(
 				expect.objectContaining({
 					region: "us-east-1",
 				}),
 			)
 
-			// Verify the endpoint property is not present
-			const clientConfig = mockBedrockRuntimeClient.mock.calls[0][0]
-			expect(clientConfig).not.toHaveProperty("endpoint")
+			const providerSettings = (mockCreateAmazonBedrock.mock.calls as unknown[][])[0][0] as Record<
+				string,
+				unknown
+			>
+			expect(providerSettings).not.toHaveProperty("baseURL")
 		})
 	})
 
 	// Test Scenario 2: Edge Case Tests
 	describe("Edge Cases", () => {
 		it("should handle empty endpoint URL gracefully", () => {
-			// Create handler with empty endpoint URL but enabled flag
 			new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test-access-key",
@@ -94,20 +94,21 @@ describe("Amazon Bedrock VPC Endpoint Functionality", () => {
 				awsBedrockEndpointEnabled: true,
 			})
 
-			// Verify the client was created without the endpoint (since it's empty)
-			expect(mockBedrockRuntimeClient).toHaveBeenCalledWith(
+			expect(mockCreateAmazonBedrock).toHaveBeenCalledWith(
 				expect.objectContaining({
 					region: "us-east-1",
 				}),
 			)
 
-			// Verify the endpoint property is not present
-			const clientConfig = mockBedrockRuntimeClient.mock.calls[0][0]
-			expect(clientConfig).not.toHaveProperty("endpoint")
+			// Empty string is falsy, so baseURL should not be set
+			const providerSettings = (mockCreateAmazonBedrock.mock.calls as unknown[][])[0][0] as Record<
+				string,
+				unknown
+			>
+			expect(providerSettings).not.toHaveProperty("baseURL")
 		})
 
 		it("should handle undefined endpoint URL gracefully", () => {
-			// Create handler with undefined endpoint URL but enabled flag
 			new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test-access-key",
@@ -117,23 +118,23 @@ describe("Amazon Bedrock VPC Endpoint Functionality", () => {
 				awsBedrockEndpointEnabled: true,
 			})
 
-			// Verify the client was created without the endpoint
-			expect(mockBedrockRuntimeClient).toHaveBeenCalledWith(
+			expect(mockCreateAmazonBedrock).toHaveBeenCalledWith(
 				expect.objectContaining({
 					region: "us-east-1",
 				}),
 			)
 
-			// Verify the endpoint property is not present
-			const clientConfig = mockBedrockRuntimeClient.mock.calls[0][0]
-			expect(clientConfig).not.toHaveProperty("endpoint")
+			const providerSettings = (mockCreateAmazonBedrock.mock.calls as unknown[][])[0][0] as Record<
+				string,
+				unknown
+			>
+			expect(providerSettings).not.toHaveProperty("baseURL")
 		})
 	})
 
-	// Test Scenario 4: Error Handling Tests
+	// Test Scenario 3: Error Handling Tests
 	describe("Error Handling", () => {
-		it("should handle invalid endpoint URLs by passing them directly to AWS SDK", () => {
-			// Create handler with an invalid URL format
+		it("should handle invalid endpoint URLs by passing them directly to the provider", () => {
 			new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test-access-key",
@@ -143,21 +144,24 @@ describe("Amazon Bedrock VPC Endpoint Functionality", () => {
 				awsBedrockEndpointEnabled: true,
 			})
 
-			// Verify the client was created with the invalid endpoint
-			// (AWS SDK will handle the validation/errors)
-			expect(mockBedrockRuntimeClient).toHaveBeenCalledWith(
+			// The invalid URL is passed directly; the provider/SDK will handle validation
+			expect(mockCreateAmazonBedrock).toHaveBeenCalledWith(
 				expect.objectContaining({
 					region: "us-east-1",
-					endpoint: "invalid-url-format",
+					baseURL: "invalid-url-format",
 				}),
 			)
 		})
 	})
 
-	// Test Scenario 5: Persistence Tests
+	// Test Scenario 4: Persistence Tests
 	describe("Persistence", () => {
 		it("should maintain consistent behavior across multiple requests", async () => {
-			// Create handler with endpoint URL and enabled flag
+			mockGenerateText.mockResolvedValue({
+				text: "test response",
+				usage: { promptTokens: 10, completionTokens: 5 },
+			})
+
 			const handler = new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test-access-key",
@@ -167,23 +171,23 @@ describe("Amazon Bedrock VPC Endpoint Functionality", () => {
 				awsBedrockEndpointEnabled: true,
 			})
 
-			// Verify the client was configured with the endpoint
-			expect(mockBedrockRuntimeClient).toHaveBeenCalledWith(
+			// Verify the provider was configured with the endpoint
+			expect(mockCreateAmazonBedrock).toHaveBeenCalledWith(
 				expect.objectContaining({
 					region: "us-east-1",
-					endpoint: "https://bedrock-vpc.example.com",
+					baseURL: "https://bedrock-vpc.example.com",
 				}),
 			)
 
 			// Make a request to ensure the endpoint configuration persists
 			try {
 				await handler.completePrompt("Test prompt")
-			} catch (error) {
-				// Ignore errors, we're just testing the client configuration persistence
+			} catch {
+				// Ignore errors — we're just testing the provider configuration persistence
 			}
 
-			// Verify the client instance was created and used
-			expect(mockBedrockRuntimeClient).toHaveBeenCalled()
+			// The provider factory should have been called exactly once (during construction)
+			expect(mockCreateAmazonBedrock).toHaveBeenCalledTimes(1)
 		})
 	})
 })

src/api/providers/__tests__/bedrock.spec.ts

@@ -18,24 +18,26 @@ vi.mock("@aws-sdk/credential-providers", () => {
 	return { fromIni: mockFromIni }
 })
 
-// Mock BedrockRuntimeClient and ConverseStreamCommand
-vi.mock("@aws-sdk/client-bedrock-runtime", () => {
-	const mockSend = vi.fn().mockResolvedValue({
-		stream: [],
-	})
-	const mockConverseStreamCommand = vi.fn()
+// Use vi.hoisted to define mock functions for AI SDK
+const { mockStreamText, mockGenerateText } = vi.hoisted(() => ({
+	mockStreamText: vi.fn(),
+	mockGenerateText: vi.fn(),
+}))
 
+vi.mock("ai", async (importOriginal) => {
+	const actual = await importOriginal<typeof import("ai")>()
 	return {
-		BedrockRuntimeClient: vi.fn().mockImplementation(() => ({
-			send: mockSend,
-		})),
-		ConverseStreamCommand: mockConverseStreamCommand,
-		ConverseCommand: vi.fn(),
+		...actual,
+		streamText: mockStreamText,
+		generateText: mockGenerateText,
 	}
 })
 
+vi.mock("@ai-sdk/amazon-bedrock", () => ({
+	createAmazonBedrock: vi.fn(() => vi.fn(() => ({ modelId: "test", provider: "bedrock" }))),
+}))
+
 import { AwsBedrockHandler } from "../bedrock"
-import { ConverseStreamCommand, BedrockRuntimeClient, ConverseCommand } from "@aws-sdk/client-bedrock-runtime"
 import {
 	BEDROCK_1M_CONTEXT_MODEL_IDS,
 	BEDROCK_SERVICE_TIER_MODEL_IDS,
@@ -45,10 +47,6 @@ import {
 
 import type { Anthropic } from "@anthropic-ai/sdk"
 
-// Get access to the mocked functions
-const mockConverseStreamCommand = vi.mocked(ConverseStreamCommand)
-const mockBedrockRuntimeClient = vi.mocked(BedrockRuntimeClient)
-
 describe("AwsBedrockHandler", () => {
 	let handler: AwsBedrockHandler
 
@@ -478,12 +476,20 @@ describe("AwsBedrockHandler", () => {
 	describe("image handling", () => {
 		const mockImageData = Buffer.from("test-image-data").toString("base64")
 
-		beforeEach(() => {
-			// Reset the mocks before each test
-			mockConverseStreamCommand.mockReset()
-		})
+		function setupMockStreamText() {
+			async function* mockFullStream() {
+				yield { type: "text-delta", text: "I see an image" }
+			}
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 10, outputTokens: 5 }),
+				providerMetadata: Promise.resolve({}),
+			})
+		}
+
+		it("should properly pass image content through to streamText via AI SDK messages", async () => {
+			setupMockStreamText()
 
-		it("should properly convert image content to Bedrock format", async () => {
 			const messages: Anthropic.Messages.MessageParam[] = [
 				{
 					role: "user",
@@ -505,42 +511,39 @@ describe("AwsBedrockHandler", () => {
 			]
 
 			const generator = handler.createMessage("", messages)
-			await generator.next() // Start the generator
-
-			// Verify the command was created with the right payload
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0]
-
-			// Verify the image was properly formatted
-			const imageBlock = commandArg.messages![0].content![0]
-			expect(imageBlock).toHaveProperty("image")
-			expect(imageBlock.image).toHaveProperty("format", "jpeg")
-			expect(imageBlock.image!.source).toHaveProperty("bytes")
-			expect(imageBlock.image!.source!.bytes).toBeInstanceOf(Uint8Array)
-		})
-
-		it("should reject unsupported image formats", async () => {
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "user",
-					content: [
-						{
-							type: "image",
-							source: {
-								type: "base64",
-								data: mockImageData,
-								media_type: "image/tiff" as "image/jpeg", // Type assertion to bypass TS
-							},
-						},
-					],
-				},
-			]
-
-			const generator = handler.createMessage("", messages)
-			await expect(generator.next()).rejects.toThrow("Unsupported image format: tiff")
+			const chunks: unknown[] = []
+			for await (const chunk of generator) {
+				chunks.push(chunk)
+			}
+
+			// Verify streamText was called
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+
+			// Verify messages were converted to AI SDK format with image parts
+			const aiSdkMessages = callArgs.messages
+			expect(aiSdkMessages).toBeDefined()
+			expect(aiSdkMessages.length).toBeGreaterThan(0)
+
+			// Find the user message containing image content
+			const userMsg = aiSdkMessages.find((m: { role: string }) => m.role === "user")
+			expect(userMsg).toBeDefined()
+			expect(Array.isArray(userMsg.content)).toBe(true)
+
+			// The AI SDK convertToAiSdkMessages converts images to { type: "image", image: "data:...", mimeType: "..." }
+			const imagePart = userMsg.content.find((p: { type: string }) => p.type === "image")
+			expect(imagePart).toBeDefined()
+			expect(imagePart.image).toContain("data:image/jpeg;base64,")
+			expect(imagePart.mimeType).toBe("image/jpeg")
+
+			const textPart = userMsg.content.find((p: { type: string }) => p.type === "text")
+			expect(textPart).toBeDefined()
+			expect(textPart.text).toBe("What's in this image?")
 		})
 
 		it("should handle multiple images in a single message", async () => {
+			setupMockStreamText()
+
 			const messages: Anthropic.Messages.MessageParam[] = [
 				{
 					role: "user",
@@ -574,20 +577,25 @@ describe("AwsBedrockHandler", () => {
 			]
 
 			const generator = handler.createMessage("", messages)
-			await generator.next() // Start the generator
-
-			// Verify the command was created with the right payload
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0]
-
-			// Verify both images were properly formatted
-			const firstImage = commandArg.messages![0].content![0]
-			const secondImage = commandArg.messages![0].content![2]
-
-			expect(firstImage).toHaveProperty("image")
-			expect(firstImage.image).toHaveProperty("format", "jpeg")
-			expect(secondImage).toHaveProperty("image")
-			expect(secondImage.image).toHaveProperty("format", "png")
+			const chunks: unknown[] = []
+			for await (const chunk of generator) {
+				chunks.push(chunk)
+			}
+
+			// Verify streamText was called
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+
+			// Verify messages contain both images
+			const userMsg = callArgs.messages.find((m: { role: string }) => m.role === "user")
+			expect(userMsg).toBeDefined()
+
+			const imageParts = userMsg.content.filter((p: { type: string }) => p.type === "image")
+			expect(imageParts).toHaveLength(2)
+			expect(imageParts[0].image).toContain("data:image/jpeg;base64,")
+			expect(imageParts[0].mimeType).toBe("image/jpeg")
+			expect(imageParts[1].image).toContain("data:image/png;base64,")
+			expect(imageParts[1].mimeType).toBe("image/png")
 		})
 	})
 
@@ -686,6 +694,17 @@ describe("AwsBedrockHandler", () => {
 	})
 
 	describe("1M context beta feature", () => {
+		function setupMockStreamText() {
+			async function* mockFullStream() {
+				yield { type: "text-delta", text: "Response" }
+			}
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 10, outputTokens: 5 }),
+				providerMetadata: Promise.resolve({}),
+			})
+		}
+
 		it("should enable 1M context window when awsBedrock1MContext is true for Claude Sonnet 4", () => {
 			const handler = new AwsBedrockHandler({
 				apiModelId: BEDROCK_1M_CONTEXT_MODEL_IDS[0],
@@ -731,7 +750,9 @@ describe("AwsBedrockHandler", () => {
 			expect(model.info.contextWindow).toBe(200_000)
 		})
 
-		it("should include anthropic_beta parameter when 1M context is enabled", async () => {
+		it("should include anthropicBeta in providerOptions when 1M context is enabled", async () => {
+			setupMockStreamText()
+
 			const handler = new AwsBedrockHandler({
 				apiModelId: BEDROCK_1M_CONTEXT_MODEL_IDS[0],
 				awsAccessKey: "test",
@@ -748,23 +769,23 @@ describe("AwsBedrockHandler", () => {
 			]
 
 			const generator = handler.createMessage("", messages)
-			await generator.next() // Start the generator
-
-			// Verify the command was created with the right payload
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
-
-			// Should include anthropic_beta in additionalModelRequestFields with both 1M context and fine-grained-tool-streaming
-			expect(commandArg.additionalModelRequestFields).toBeDefined()
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).toContain("context-1m-2025-08-07")
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).toContain(
-				"fine-grained-tool-streaming-2025-05-14",
-			)
-			// Should not include anthropic_version since thinking is not enabled
-			expect(commandArg.additionalModelRequestFields.anthropic_version).toBeUndefined()
+			const chunks: unknown[] = []
+			for await (const chunk of generator) {
+				chunks.push(chunk)
+			}
+
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+
+			// Should include anthropicBeta in providerOptions.bedrock with 1M context
+			const bedrockOpts = callArgs.providerOptions?.bedrock as Record<string, unknown> | undefined
+			expect(bedrockOpts).toBeDefined()
+			expect(bedrockOpts!.anthropicBeta).toContain("context-1m-2025-08-07")
 		})
 
-		it("should not include 1M context beta when 1M context is disabled but still include fine-grained-tool-streaming", async () => {
+		it("should not include 1M context beta when 1M context is disabled", async () => {
+			setupMockStreamText()
+
 			const handler = new AwsBedrockHandler({
 				apiModelId: BEDROCK_1M_CONTEXT_MODEL_IDS[0],
 				awsAccessKey: "test",
@@ -781,22 +802,24 @@ describe("AwsBedrockHandler", () => {
 			]
 
 			const generator = handler.createMessage("", messages)
-			await generator.next() // Start the generator
-
-			// Verify the command was created with the right payload
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
-
-			// Should include anthropic_beta with fine-grained-tool-streaming for Claude models
-			expect(commandArg.additionalModelRequestFields).toBeDefined()
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).toContain(
-				"fine-grained-tool-streaming-2025-05-14",
-			)
-			// Should NOT include 1M context beta
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).not.toContain("context-1m-2025-08-07")
+			const chunks: unknown[] = []
+			for await (const chunk of generator) {
+				chunks.push(chunk)
+			}
+
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+
+			// Should NOT include anthropicBeta with 1M context
+			const bedrockOpts = callArgs.providerOptions?.bedrock as Record<string, unknown> | undefined
+			if (bedrockOpts?.anthropicBeta) {
+				expect(bedrockOpts.anthropicBeta).not.toContain("context-1m-2025-08-07")
+			}
 		})
 
-		it("should not include 1M context beta for non-Claude Sonnet 4 models but still include fine-grained-tool-streaming", async () => {
+		it("should not include 1M context beta for non-Claude Sonnet 4 models", async () => {
+			setupMockStreamText()
+
 			const handler = new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test",
@@ -813,19 +836,19 @@ describe("AwsBedrockHandler", () => {
 			]
 
 			const generator = handler.createMessage("", messages)
-			await generator.next() // Start the generator
-
-			// Verify the command was created with the right payload
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
-
-			// Should include anthropic_beta with fine-grained-tool-streaming for Claude models (even non-Sonnet 4)
-			expect(commandArg.additionalModelRequestFields).toBeDefined()
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).toContain(
-				"fine-grained-tool-streaming-2025-05-14",
-			)
-			// Should NOT include 1M context beta for non-Sonnet 4 models
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).not.toContain("context-1m-2025-08-07")
+			const chunks: unknown[] = []
+			for await (const chunk of generator) {
+				chunks.push(chunk)
+			}
+
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+
+			// Should NOT include anthropicBeta with 1M context for non-Sonnet 4 models
+			const bedrockOpts = callArgs.providerOptions?.bedrock as Record<string, unknown> | undefined
+			if (bedrockOpts?.anthropicBeta) {
+				expect(bedrockOpts.anthropicBeta).not.toContain("context-1m-2025-08-07")
+			}
 		})
 
 		it("should enable 1M context window with cross-region inference for Claude Sonnet 4", () => {
@@ -846,7 +869,9 @@ describe("AwsBedrockHandler", () => {
 			expect(model.id).toBe(`us.${BEDROCK_1M_CONTEXT_MODEL_IDS[0]}`)
 		})
 
-		it("should include anthropic_beta parameter with cross-region inference for Claude Sonnet 4", async () => {
+		it("should include anthropicBeta with cross-region inference for Claude Sonnet 4", async () => {
+			setupMockStreamText()
+
 			const handler = new AwsBedrockHandler({
 				apiModelId: BEDROCK_1M_CONTEXT_MODEL_IDS[0],
 				awsAccessKey: "test",
@@ -864,33 +889,34 @@ describe("AwsBedrockHandler", () => {
 			]
 
 			const generator = handler.createMessage("", messages)
-			await generator.next() // Start the generator
-
-			// Verify the command was created with the right payload
-			expect(mockConverseStreamCommand).toHaveBeenCalled()
-			const commandArg = mockConverseStreamCommand.mock.calls[
-				mockConverseStreamCommand.mock.calls.length - 1
-			][0] as any
-
-			// Should include anthropic_beta in additionalModelRequestFields with both 1M context and fine-grained-tool-streaming
-			expect(commandArg.additionalModelRequestFields).toBeDefined()
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).toContain("context-1m-2025-08-07")
-			expect(commandArg.additionalModelRequestFields.anthropic_beta).toContain(
-				"fine-grained-tool-streaming-2025-05-14",
-			)
-			// Should not include anthropic_version since thinking is not enabled
-			expect(commandArg.additionalModelRequestFields.anthropic_version).toBeUndefined()
-			// Model ID should have cross-region prefix
-			expect(commandArg.modelId).toBe(`us.${BEDROCK_1M_CONTEXT_MODEL_IDS[0]}`)
+			const chunks: unknown[] = []
+			for await (const chunk of generator) {
+				chunks.push(chunk)
+			}
+
+			expect(mockStreamText).toHaveBeenCalledTimes(1)
+			const callArgs = mockStreamText.mock.calls[0][0]
+
+			// Should include anthropicBeta in providerOptions.bedrock with 1M context
+			const bedrockOpts = callArgs.providerOptions?.bedrock as Record<string, unknown> | undefined
+			expect(bedrockOpts).toBeDefined()
+			expect(bedrockOpts!.anthropicBeta).toContain("context-1m-2025-08-07")
 		})
 	})
 
 	describe("service tier feature", () => {
 		const supportedModelId = BEDROCK_SERVICE_TIER_MODEL_IDS[0] // amazon.nova-lite-v1:0
 
-		beforeEach(() => {
-			mockConverseStreamCommand.mockReset()
-		})
+		function setupMockStreamText() {
+			async function* mockFullStream() {
+				yield { type: "text-delta", text: "Response" }
+			}
+			mockStreamText.mockReturnValue({
+				fullStream: mockFullStream(),
+				usage: Promise.resolve({ inputTokens: 10, outputTokens: 5 }),
+				providerMetadata: Promise.resolve({}),
+			})
+		}
 
 		describe("pricing multipliers in getModel()", () => {
 			it("should apply FLEX tier pricing with 50% discount", () => {
@@ -976,7 +1002,9 @@ describe("AwsBedrockHandler", () => {
 		})
 
 		describe("service_tier parameter in API requests", () => {
-			it("should include service_tier as top-level parameter for supported models", async () => {
+			it("should include service_tier in providerOptions.bedrock.additionalModelRequestFields for supported models", async () => {
+				setupMockStreamText()
+
 				const handler = new AwsBedrockHandler({
 					apiModelId: supportedModelId,
 					awsAccessKey: "test",
@@ -993,23 +1021,27 @@ describe("AwsBedrockHandler", () => {
 				]
 
 				const generator = handler.createMessage("", messages)
-				await generator.next() // Start the generator
-
-				// Verify the command was created with service_tier at top level
-				// Per AWS documentation, service_tier must be a top-level parameter, not inside additionalModelRequestFields
-				// https://docs.aws.amazon.com/bedrock/latest/userguide/service-tiers-inference.html
-				expect(mockConverseStreamCommand).toHaveBeenCalled()
-				const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
-
-				// service_tier should be at the top level of the payload
-				expect(commandArg.service_tier).toBe("PRIORITY")
-				// service_tier should NOT be in additionalModelRequestFields
-				if (commandArg.additionalModelRequestFields) {
-					expect(commandArg.additionalModelRequestFields.service_tier).toBeUndefined()
+				const chunks: unknown[] = []
+				for await (const chunk of generator) {
+					chunks.push(chunk)
 				}
+
+				expect(mockStreamText).toHaveBeenCalledTimes(1)
+				const callArgs = mockStreamText.mock.calls[0][0]
+
+				// service_tier should be passed through providerOptions.bedrock.additionalModelRequestFields
+				const bedrockOpts = callArgs.providerOptions?.bedrock as Record<string, unknown> | undefined
+				expect(bedrockOpts).toBeDefined()
+				const additionalFields = bedrockOpts!.additionalModelRequestFields as
+					| Record<string, unknown>
+					| undefined
+				expect(additionalFields).toBeDefined()
+				expect(additionalFields!.service_tier).toBe("PRIORITY")
 			})
 
-			it("should include service_tier FLEX as top-level parameter", async () => {
+			it("should include service_tier FLEX in providerOptions", async () => {
+				setupMockStreamText()
+
 				const handler = new AwsBedrockHandler({
 					apiModelId: supportedModelId,
 					awsAccessKey: "test",
@@ -1026,20 +1058,26 @@ describe("AwsBedrockHandler", () => {
 				]
 
 				const generator = handler.createMessage("", messages)
-				await generator.next() // Start the generator
+				const chunks: unknown[] = []
+				for await (const chunk of generator) {
+					chunks.push(chunk)
+				}
 
-				expect(mockConverseStreamCommand).toHaveBeenCalled()
-				const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
+				expect(mockStreamText).toHaveBeenCalledTimes(1)
+				const callArgs = mockStreamText.mock.calls[0][0]
 
-				// service_tier should be at the top level of the payload
-				expect(commandArg.service_tier).toBe("FLEX")
-				// service_tier should NOT be in additionalModelRequestFields
-				if (commandArg.additionalModelRequestFields) {
-					expect(commandArg.additionalModelRequestFields.service_tier).toBeUndefined()
-				}
+				const bedrockOpts = callArgs.providerOptions?.bedrock as Record<string, unknown> | undefined
+				expect(bedrockOpts).toBeDefined()
+				const additionalFields = bedrockOpts!.additionalModelRequestFields as
+					| Record<string, unknown>
+					| undefined
+				expect(additionalFields).toBeDefined()
+				expect(additionalFields!.service_tier).toBe("FLEX")
 			})
 
 			it("should NOT include service_tier for unsupported models", async () => {
+				setupMockStreamText()
+
 				const unsupportedModelId = "anthropic.claude-3-5-sonnet-20241022-v2:0"
 				const handler = new AwsBedrockHandler({
 					apiModelId: unsupportedModelId,
@@ -1057,19 +1095,25 @@ describe("AwsBedrockHandler", () => {
 				]
 
 				const generator = handler.createMessage("", messages)
-				await generator.next() // Start the generator
+				const chunks: unknown[] = []
+				for await (const chunk of generator) {
+					chunks.push(chunk)
+				}
 
-				expect(mockConverseStreamCommand).toHaveBeenCalled()
-				const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
+				expect(mockStreamText).toHaveBeenCalledTimes(1)
+				const callArgs = mockStreamText.mock.calls[0][0]
 
-				// Service tier should NOT be included for unsupported models (at top level or in additionalModelRequestFields)
-				expect(commandArg.service_tier).toBeUndefined()
-				if (commandArg.additionalModelRequestFields) {
-					expect(commandArg.additionalModelRequestFields.service_tier).toBeUndefined()
+				// Service tier should NOT be included for unsupported models
+				const bedrockOpts = callArgs.providerOptions?.bedrock as Record<string, unknown> | undefined
+				if (bedrockOpts?.additionalModelRequestFields) {
+					const additionalFields = bedrockOpts.additionalModelRequestFields as Record<string, unknown>
+					expect(additionalFields.service_tier).toBeUndefined()
 				}
 			})
 
 			it("should NOT include service_tier when not specified", async () => {
+				setupMockStreamText()
+
 				const handler = new AwsBedrockHandler({
 					apiModelId: supportedModelId,
 					awsAccessKey: "test",
@@ -1086,15 +1130,19 @@ describe("AwsBedrockHandler", () => {
 				]
 
 				const generator = handler.createMessage("", messages)
-				await generator.next() // Start the generator
+				const chunks: unknown[] = []
+				for await (const chunk of generator) {
+					chunks.push(chunk)
+				}
 
-				expect(mockConverseStreamCommand).toHaveBeenCalled()
-				const commandArg = mockConverseStreamCommand.mock.calls[0][0] as any
+				expect(mockStreamText).toHaveBeenCalledTimes(1)
+				const callArgs = mockStreamText.mock.calls[0][0]
 
-				// Service tier should NOT be included when not specified (at top level or in additionalModelRequestFields)
-				expect(commandArg.service_tier).toBeUndefined()
-				if (commandArg.additionalModelRequestFields) {
-					expect(commandArg.additionalModelRequestFields.service_tier).toBeUndefined()
+				// Service tier should NOT be included when not specified
+				const bedrockOpts = callArgs.providerOptions?.bedrock as Record<string, unknown> | undefined
+				if (bedrockOpts?.additionalModelRequestFields) {
+					const additionalFields = bedrockOpts.additionalModelRequestFields as Record<string, unknown>
+					expect(additionalFields.service_tier).toBeUndefined()
 				}
 			})
 		})
@@ -1127,16 +1175,16 @@ describe("AwsBedrockHandler", () => {
 	})
 
 	describe("error telemetry", () => {
-		let mockSend: ReturnType<typeof vi.fn>
-
 		beforeEach(() => {
 			mockCaptureException.mockClear()
-			// Get access to the mock send function from the mocked client
-			mockSend = vi.mocked(BedrockRuntimeClient).mock.results[0]?.value?.send
 		})
 
 		it("should capture telemetry on createMessage error", async () => {
-			// Create a handler with a fresh mock
+			// Mock streamText to throw an error
+			mockStreamText.mockImplementation(() => {
+				throw new Error("Bedrock API error")
+			})
+
 			const errorHandler = new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test-access-key",
@@ -1144,15 +1192,6 @@ describe("AwsBedrockHandler", () => {
 				awsRegion: "us-east-1",
 			})
 
-			// Get the mock send from the new handler instance
-			const clientInstance =
-				vi.mocked(BedrockRuntimeClient).mock.results[vi.mocked(BedrockRuntimeClient).mock.results.length - 1]
-					?.value
-			const mockSendFn = clientInstance?.send as ReturnType<typeof vi.fn>
-
-			// Mock the send to throw an error
-			mockSendFn.mockRejectedValueOnce(new Error("Bedrock API error"))
-
 			const messages: Anthropic.Messages.MessageParam[] = [
 				{
 					role: "user",
@@ -1186,7 +1225,9 @@ describe("AwsBedrockHandler", () => {
 		})
 
 		it("should capture telemetry on completePrompt error", async () => {
-			// Create a handler with a fresh mock
+			// Mock generateText to throw an error
+			mockGenerateText.mockRejectedValueOnce(new Error("Bedrock completion error"))
+
 			const errorHandler = new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test-access-key",
@@ -1194,15 +1235,6 @@ describe("AwsBedrockHandler", () => {
 				awsRegion: "us-east-1",
 			})
 
-			// Get the mock send from the new handler instance
-			const clientInstance =
-				vi.mocked(BedrockRuntimeClient).mock.results[vi.mocked(BedrockRuntimeClient).mock.results.length - 1]
-					?.value
-			const mockSendFn = clientInstance?.send as ReturnType<typeof vi.fn>
-
-			// Mock the send to throw an error for ConverseCommand
-			mockSendFn.mockRejectedValueOnce(new Error("Bedrock completion error"))
-
 			// Call completePrompt - it should throw
 			await expect(errorHandler.completePrompt("Test prompt")).rejects.toThrow()
 
@@ -1223,7 +1255,11 @@ describe("AwsBedrockHandler", () => {
 		})
 
 		it("should still throw the error after capturing telemetry", async () => {
-			// Create a handler with a fresh mock
+			// Mock streamText to throw an error
+			mockStreamText.mockImplementation(() => {
+				throw new Error("Test error for throw verification")
+			})
+
 			const errorHandler = new AwsBedrockHandler({
 				apiModelId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
 				awsAccessKey: "test-access-key",
@@ -1231,15 +1267,6 @@ describe("AwsBedrockHandler", () => {
 				awsRegion: "us-east-1",
 			})
 
-			// Get the mock send from the new handler instance
-			const clientInstance =
-				vi.mocked(BedrockRuntimeClient).mock.results[vi.mocked(BedrockRuntimeClient).mock.results.length - 1]
-					?.value
-			const mockSendFn = clientInstance?.send as ReturnType<typeof vi.fn>
-
-			// Mock the send to throw an error
-			mockSendFn.mockRejectedValueOnce(new Error("Test error for throw verification"))
-
 			const messages: Anthropic.Messages.MessageParam[] = [
 				{
 					role: "user",

src/api/transform/__tests__/ai-sdk.spec.ts

@@ -349,7 +349,14 @@ describe("AI SDK conversion utilities", () => {
 			expect(result[0]).toEqual({
 				role: "assistant",
 				content: [
-					{ type: "reasoning", text: "Deep thought" },
+					{
+						type: "reasoning",
+						text: "Deep thought",
+						providerOptions: {
+							bedrock: { signature: "sig" },
+							anthropic: { signature: "sig" },
+						},
+					},
 					{ type: "text", text: "OK" },
 				],
 			})

src/api/transform/__tests__/bedrock-converse-format.spec.ts

@@ -1,694 +0,0 @@
-// npx vitest run src/api/transform/__tests__/bedrock-converse-format.spec.ts
-
-import { convertToBedrockConverseMessages } from "../bedrock-converse-format"
-import { Anthropic } from "@anthropic-ai/sdk"
-import { ContentBlock, ToolResultContentBlock } from "@aws-sdk/client-bedrock-runtime"
-import { OPENAI_CALL_ID_MAX_LENGTH } from "../../../utils/tool-id"
-
-describe("convertToBedrockConverseMessages", () => {
-	it("converts simple text messages correctly", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{ role: "user", content: "Hello" },
-			{ role: "assistant", content: "Hi there" },
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		expect(result).toEqual([
-			{
-				role: "user",
-				content: [{ text: "Hello" }],
-			},
-			{
-				role: "assistant",
-				content: [{ text: "Hi there" }],
-			},
-		])
-	})
-
-	it("converts messages with images correctly", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "user",
-				content: [
-					{
-						type: "text",
-						text: "Look at this image:",
-					},
-					{
-						type: "image",
-						source: {
-							type: "base64",
-							data: "SGVsbG8=", // "Hello" in base64
-							media_type: "image/jpeg" as const,
-						},
-					},
-				],
-			},
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		if (!result[0] || !result[0].content) {
-			expect.fail("Expected result to have content")
-			return
-		}
-
-		expect(result[0].role).toBe("user")
-		expect(result[0].content).toHaveLength(2)
-		expect(result[0].content[0]).toEqual({ text: "Look at this image:" })
-
-		const imageBlock = result[0].content[1] as ContentBlock
-		if ("image" in imageBlock && imageBlock.image && imageBlock.image.source) {
-			expect(imageBlock.image.format).toBe("jpeg")
-			expect(imageBlock.image.source).toBeDefined()
-			expect(imageBlock.image.source.bytes).toBeDefined()
-		} else {
-			expect.fail("Expected image block not found")
-		}
-	})
-
-	it("converts tool use messages correctly (native tools format; default)", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "assistant",
-				content: [
-					{
-						type: "tool_use",
-						id: "test-id",
-						name: "read_file",
-						input: {
-							path: "test.txt",
-						},
-					},
-				],
-			},
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		if (!result[0] || !result[0].content) {
-			expect.fail("Expected result to have content")
-			return
-		}
-
-		expect(result[0].role).toBe("assistant")
-		const toolBlock = result[0].content[0] as ContentBlock
-		if ("toolUse" in toolBlock && toolBlock.toolUse) {
-			expect(toolBlock.toolUse).toEqual({
-				toolUseId: "test-id",
-				name: "read_file",
-				input: { path: "test.txt" },
-			})
-		} else {
-			expect.fail("Expected tool use block not found")
-		}
-	})
-
-	it("converts tool use messages correctly (native tools format)", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "assistant",
-				content: [
-					{
-						type: "tool_use",
-						id: "test-id",
-						name: "read_file",
-						input: {
-							path: "test.txt",
-						},
-					},
-				],
-			},
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		if (!result[0] || !result[0].content) {
-			expect.fail("Expected result to have content")
-			return
-		}
-
-		expect(result[0].role).toBe("assistant")
-		const toolBlock = result[0].content[0] as ContentBlock
-		if ("toolUse" in toolBlock && toolBlock.toolUse) {
-			expect(toolBlock.toolUse).toEqual({
-				toolUseId: "test-id",
-				name: "read_file",
-				input: { path: "test.txt" },
-			})
-		} else {
-			expect.fail("Expected tool use block not found")
-		}
-	})
-
-	it("converts tool result messages to native format (default)", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "user",
-				content: [
-					{
-						type: "tool_result",
-						tool_use_id: "test-id",
-						content: [{ type: "text", text: "File contents here" }],
-					},
-				],
-			},
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		if (!result[0] || !result[0].content) {
-			expect.fail("Expected result to have content")
-			return
-		}
-
-		expect(result[0].role).toBe("user")
-		const resultBlock = result[0].content[0] as ContentBlock
-		if ("toolResult" in resultBlock && resultBlock.toolResult) {
-			const expectedContent: ToolResultContentBlock[] = [{ text: "File contents here" }]
-			expect(resultBlock.toolResult).toEqual({
-				toolUseId: "test-id",
-				content: expectedContent,
-				status: "success",
-			})
-		} else {
-			expect.fail("Expected tool result block not found")
-		}
-	})
-
-	it("converts tool result messages to native format", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "user",
-				content: [
-					{
-						type: "tool_result",
-						tool_use_id: "test-id",
-						content: [{ type: "text", text: "File contents here" }],
-					},
-				],
-			},
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		if (!result[0] || !result[0].content) {
-			expect.fail("Expected result to have content")
-			return
-		}
-
-		expect(result[0].role).toBe("user")
-		const resultBlock = result[0].content[0] as ContentBlock
-		if ("toolResult" in resultBlock && resultBlock.toolResult) {
-			const expectedContent: ToolResultContentBlock[] = [{ text: "File contents here" }]
-			expect(resultBlock.toolResult).toEqual({
-				toolUseId: "test-id",
-				content: expectedContent,
-				status: "success",
-			})
-		} else {
-			expect.fail("Expected tool result block not found")
-		}
-	})
-
-	it("converts tool result messages with string content to native format (default)", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "user",
-				content: [
-					{
-						type: "tool_result",
-						tool_use_id: "test-id",
-						content: "File: test.txt\nLines 1-5:\nHello World",
-					} as any, // Anthropic types don't allow string content but runtime can have it
-				],
-			},
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		if (!result[0] || !result[0].content) {
-			expect.fail("Expected result to have content")
-			return
-		}
-
-		expect(result[0].role).toBe("user")
-		const resultBlock = result[0].content[0] as ContentBlock
-		if ("toolResult" in resultBlock && resultBlock.toolResult) {
-			expect(resultBlock.toolResult).toEqual({
-				toolUseId: "test-id",
-				content: [{ text: "File: test.txt\nLines 1-5:\nHello World" }],
-				status: "success",
-			})
-		} else {
-			expect.fail("Expected tool result block not found")
-		}
-	})
-
-	it("converts tool result messages with string content to native format", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "user",
-				content: [
-					{
-						type: "tool_result",
-						tool_use_id: "test-id",
-						content: "File: test.txt\nLines 1-5:\nHello World",
-					} as any, // Anthropic types don't allow string content but runtime can have it
-				],
-			},
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		if (!result[0] || !result[0].content) {
-			expect.fail("Expected result to have content")
-			return
-		}
-
-		expect(result[0].role).toBe("user")
-		const resultBlock = result[0].content[0] as ContentBlock
-		if ("toolResult" in resultBlock && resultBlock.toolResult) {
-			expect(resultBlock.toolResult).toEqual({
-				toolUseId: "test-id",
-				content: [{ text: "File: test.txt\nLines 1-5:\nHello World" }],
-				status: "success",
-			})
-		} else {
-			expect.fail("Expected tool result block not found")
-		}
-	})
-
-	it("keeps both tool_use and tool_result in native format by default", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "assistant",
-				content: [
-					{
-						type: "tool_use",
-						id: "call-123",
-						name: "read_file",
-						input: { path: "test.txt" },
-					},
-				],
-			},
-			{
-				role: "user",
-				content: [
-					{
-						type: "tool_result",
-						tool_use_id: "call-123",
-						content: "File contents here",
-					} as any,
-				],
-			},
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		// Both should be native toolUse/toolResult blocks
-		const assistantContent = result[0]?.content?.[0] as ContentBlock
-		const userContent = result[1]?.content?.[0] as ContentBlock
-
-		expect("toolUse" in assistantContent).toBe(true)
-		expect("toolResult" in userContent).toBe(true)
-		expect("text" in assistantContent).toBe(false)
-		expect("text" in userContent).toBe(false)
-	})
-
-	it("handles text content correctly", () => {
-		const messages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "user",
-				content: [
-					{
-						type: "text",
-						text: "Hello world",
-					},
-				],
-			},
-		]
-
-		const result = convertToBedrockConverseMessages(messages)
-
-		if (!result[0] || !result[0].content) {
-			expect.fail("Expected result to have content")
-			return
-		}
-
-		expect(result[0].role).toBe("user")
-		expect(result[0].content).toHaveLength(1)
-		const textBlock = result[0].content[0] as ContentBlock
-		expect(textBlock).toEqual({ text: "Hello world" })
-	})
-
-	describe("toolUseId sanitization for Bedrock 64-char limit", () => {
-		it("truncates toolUseId longer than 64 characters in tool_use blocks", () => {
-			const longId = "a".repeat(100)
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [
-						{
-							type: "tool_use",
-							id: longId,
-							name: "read_file",
-							input: { path: "test.txt" },
-						},
-					],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-			const toolBlock = result[0]?.content?.[0] as ContentBlock
-
-			if ("toolUse" in toolBlock && toolBlock.toolUse && toolBlock.toolUse.toolUseId) {
-				expect(toolBlock.toolUse.toolUseId.length).toBeLessThanOrEqual(OPENAI_CALL_ID_MAX_LENGTH)
-				expect(toolBlock.toolUse.toolUseId.length).toBe(OPENAI_CALL_ID_MAX_LENGTH)
-				expect(toolBlock.toolUse.toolUseId).toContain("_")
-			} else {
-				expect.fail("Expected tool use block not found")
-			}
-		})
-
-		it("truncates toolUseId longer than 64 characters in tool_result blocks with string content", () => {
-			const longId = "b".repeat(100)
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "user",
-					content: [
-						{
-							type: "tool_result",
-							tool_use_id: longId,
-							content: "Result content",
-						} as any,
-					],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-			const resultBlock = result[0]?.content?.[0] as ContentBlock
-
-			if ("toolResult" in resultBlock && resultBlock.toolResult && resultBlock.toolResult.toolUseId) {
-				expect(resultBlock.toolResult.toolUseId.length).toBeLessThanOrEqual(OPENAI_CALL_ID_MAX_LENGTH)
-				expect(resultBlock.toolResult.toolUseId.length).toBe(OPENAI_CALL_ID_MAX_LENGTH)
-				expect(resultBlock.toolResult.toolUseId).toContain("_")
-			} else {
-				expect.fail("Expected tool result block not found")
-			}
-		})
-
-		it("truncates toolUseId longer than 64 characters in tool_result blocks with array content", () => {
-			const longId = "c".repeat(100)
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "user",
-					content: [
-						{
-							type: "tool_result",
-							tool_use_id: longId,
-							content: [{ type: "text", text: "Result content" }],
-						},
-					],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-			const resultBlock = result[0]?.content?.[0] as ContentBlock
-
-			if ("toolResult" in resultBlock && resultBlock.toolResult && resultBlock.toolResult.toolUseId) {
-				expect(resultBlock.toolResult.toolUseId.length).toBeLessThanOrEqual(OPENAI_CALL_ID_MAX_LENGTH)
-				expect(resultBlock.toolResult.toolUseId.length).toBe(OPENAI_CALL_ID_MAX_LENGTH)
-			} else {
-				expect.fail("Expected tool result block not found")
-			}
-		})
-
-		it("keeps toolUseId unchanged when under 64 characters", () => {
-			const shortId = "short-id-123"
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [
-						{
-							type: "tool_use",
-							id: shortId,
-							name: "read_file",
-							input: { path: "test.txt" },
-						},
-					],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-			const toolBlock = result[0]?.content?.[0] as ContentBlock
-
-			if ("toolUse" in toolBlock && toolBlock.toolUse) {
-				expect(toolBlock.toolUse.toolUseId).toBe(shortId)
-			} else {
-				expect.fail("Expected tool use block not found")
-			}
-		})
-
-		it("produces consistent truncated IDs for the same input", () => {
-			const longId = "d".repeat(100)
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [
-						{
-							type: "tool_use",
-							id: longId,
-							name: "read_file",
-							input: { path: "test.txt" },
-						},
-					],
-				},
-			]
-
-			const result1 = convertToBedrockConverseMessages(messages)
-			const result2 = convertToBedrockConverseMessages(messages)
-
-			const toolBlock1 = result1[0]?.content?.[0] as ContentBlock
-			const toolBlock2 = result2[0]?.content?.[0] as ContentBlock
-
-			if ("toolUse" in toolBlock1 && toolBlock1.toolUse && "toolUse" in toolBlock2 && toolBlock2.toolUse) {
-				expect(toolBlock1.toolUse.toolUseId).toBe(toolBlock2.toolUse.toolUseId)
-			} else {
-				expect.fail("Expected tool use blocks not found")
-			}
-		})
-
-		it("produces different truncated IDs for different long inputs", () => {
-			const longId1 = "e".repeat(100)
-			const longId2 = "f".repeat(100)
-
-			const messages1: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [{ type: "tool_use", id: longId1, name: "read_file", input: {} }],
-				},
-			]
-			const messages2: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [{ type: "tool_use", id: longId2, name: "read_file", input: {} }],
-				},
-			]
-
-			const result1 = convertToBedrockConverseMessages(messages1)
-			const result2 = convertToBedrockConverseMessages(messages2)
-
-			const toolBlock1 = result1[0]?.content?.[0] as ContentBlock
-			const toolBlock2 = result2[0]?.content?.[0] as ContentBlock
-
-			if ("toolUse" in toolBlock1 && toolBlock1.toolUse && "toolUse" in toolBlock2 && toolBlock2.toolUse) {
-				expect(toolBlock1.toolUse.toolUseId).not.toBe(toolBlock2.toolUse.toolUseId)
-			} else {
-				expect.fail("Expected tool use blocks not found")
-			}
-		})
-
-		it("matching tool_use and tool_result IDs are both truncated consistently", () => {
-			const longId = "g".repeat(100)
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [
-						{
-							type: "tool_use",
-							id: longId,
-							name: "read_file",
-							input: { path: "test.txt" },
-						},
-					],
-				},
-				{
-					role: "user",
-					content: [
-						{
-							type: "tool_result",
-							tool_use_id: longId,
-							content: "File contents",
-						} as any,
-					],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-
-			const toolUseBlock = result[0]?.content?.[0] as ContentBlock
-			const toolResultBlock = result[1]?.content?.[0] as ContentBlock
-
-			if (
-				"toolUse" in toolUseBlock &&
-				toolUseBlock.toolUse &&
-				toolUseBlock.toolUse.toolUseId &&
-				"toolResult" in toolResultBlock &&
-				toolResultBlock.toolResult &&
-				toolResultBlock.toolResult.toolUseId
-			) {
-				expect(toolUseBlock.toolUse.toolUseId).toBe(toolResultBlock.toolResult.toolUseId)
-				expect(toolUseBlock.toolUse.toolUseId.length).toBeLessThanOrEqual(OPENAI_CALL_ID_MAX_LENGTH)
-			} else {
-				expect.fail("Expected tool use and result blocks not found")
-			}
-		})
-	})
-
-	describe("thinking and reasoning block handling", () => {
-		it("should convert thinking blocks to reasoningContent format", () => {
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [
-						{ type: "thinking", thinking: "Let me think about this...", signature: "sig-abc123" } as any,
-						{ type: "text", text: "Here is my answer." },
-					],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-
-			expect(result).toHaveLength(1)
-			expect(result[0].role).toBe("assistant")
-			expect(result[0].content).toHaveLength(2)
-
-			const reasoningBlock = result[0].content![0] as any
-			expect(reasoningBlock.reasoningContent).toBeDefined()
-			expect(reasoningBlock.reasoningContent.reasoningText.text).toBe("Let me think about this...")
-			expect(reasoningBlock.reasoningContent.reasoningText.signature).toBe("sig-abc123")
-
-			const textBlock = result[0].content![1] as any
-			expect(textBlock.text).toBe("Here is my answer.")
-		})
-
-		it("should convert redacted_thinking blocks with data to reasoningContent.redactedContent", () => {
-			const testData = Buffer.from("encrypted-redacted-content").toString("base64")
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [{ type: "redacted_thinking", data: testData } as any, { type: "text", text: "Response" }],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-
-			expect(result).toHaveLength(1)
-			expect(result[0].content).toHaveLength(2)
-
-			const redactedBlock = result[0].content![0] as any
-			expect(redactedBlock.reasoningContent).toBeDefined()
-			expect(redactedBlock.reasoningContent.redactedContent).toBeInstanceOf(Uint8Array)
-			// Verify round-trip: decode back and compare
-			const decoded = Buffer.from(redactedBlock.reasoningContent.redactedContent).toString("utf-8")
-			expect(decoded).toBe("encrypted-redacted-content")
-		})
-
-		it("should skip redacted_thinking blocks without data", () => {
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [{ type: "redacted_thinking" } as any, { type: "text", text: "Response" }],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-
-			expect(result).toHaveLength(1)
-			// Only the text block should remain (redacted_thinking without data is filtered out)
-			expect(result[0].content).toHaveLength(1)
-			expect((result[0].content![0] as any).text).toBe("Response")
-		})
-
-		it("should skip reasoning blocks (internal Roo Code format)", () => {
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [
-						{ type: "reasoning", text: "Internal reasoning" } as any,
-						{ type: "text", text: "Response" },
-					],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-
-			expect(result).toHaveLength(1)
-			expect(result[0].content).toHaveLength(1)
-			expect((result[0].content![0] as any).text).toBe("Response")
-		})
-
-		it("should skip thoughtSignature blocks (Gemini format)", () => {
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [
-						{ type: "text", text: "Response" },
-						{ type: "thoughtSignature", thoughtSignature: "gemini-sig" } as any,
-					],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-
-			expect(result).toHaveLength(1)
-			expect(result[0].content).toHaveLength(1)
-			expect((result[0].content![0] as any).text).toBe("Response")
-		})
-
-		it("should handle full thinking + redacted_thinking + text + tool_use message", () => {
-			const redactedData = Buffer.from("redacted-binary").toString("base64")
-			const messages: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "assistant",
-					content: [
-						{ type: "thinking", thinking: "Deep thought", signature: "sig-xyz" } as any,
-						{ type: "redacted_thinking", data: redactedData } as any,
-						{ type: "text", text: "I'll use a tool." },
-						{ type: "tool_use", id: "tool-1", name: "read_file", input: { path: "test.txt" } },
-					],
-				},
-			]
-
-			const result = convertToBedrockConverseMessages(messages)
-
-			expect(result).toHaveLength(1)
-			expect(result[0].content).toHaveLength(4)
-
-			// thinking → reasoningContent.reasoningText
-			expect((result[0].content![0] as any).reasoningContent.reasoningText.text).toBe("Deep thought")
-			expect((result[0].content![0] as any).reasoningContent.reasoningText.signature).toBe("sig-xyz")
-
-			// redacted_thinking → reasoningContent.redactedContent
-			expect((result[0].content![1] as any).reasoningContent.redactedContent).toBeInstanceOf(Uint8Array)
-
-			// text
-			expect((result[0].content![2] as any).text).toBe("I'll use a tool.")
-
-			// tool_use → toolUse
-			expect((result[0].content![3] as any).toolUse.name).toBe("read_file")
-		})
-	})
-})

src/api/transform/ai-sdk.ts

@@ -139,6 +139,11 @@ export function convertToAiSdkMessages(
 					providerOptions?: Record<string, Record<string, unknown>>
 				}> = []
 
+				// Capture thinking signature for Anthropic-protocol providers (Bedrock, Anthropic).
+				// Task.ts stores thinking blocks as { type: "thinking", thinking: "...", signature: "..." }.
+				// The signature must be passed back via providerOptions on reasoning parts.
+				let thinkingSignature: string | undefined
+
 				// Extract thoughtSignature from content blocks (Gemini 3 thought signature round-tripping).
 				// Task.ts stores these as { type: "thoughtSignature", thoughtSignature: "..." } blocks.
 				let thoughtSignature: string | undefined
@@ -196,16 +201,20 @@ export function convertToAiSdkMessages(
 					if ((part as unknown as { type?: string }).type === "thinking") {
 						if (reasoningContent) continue
 
-						const thinking = (part as unknown as { thinking?: string }).thinking
-						if (typeof thinking === "string" && thinking.length > 0) {
-							reasoningParts.push(thinking)
+						const thinkingPart = part as unknown as { thinking?: string; signature?: string }
+						if (typeof thinkingPart.thinking === "string" && thinkingPart.thinking.length > 0) {
+							reasoningParts.push(thinkingPart.thinking)
+						}
+						// Capture the signature for round-tripping (Anthropic/Bedrock thinking)
+						if (thinkingPart.signature) {
+							thinkingSignature = thinkingPart.signature
 						}
 						continue
 					}
 				}
 
 				const content: Array<
-					| { type: "reasoning"; text: string }
+					| { type: "reasoning"; text: string; providerOptions?: Record<string, Record<string, unknown>> }
 					| { type: "text"; text: string }
 					| {
 							type: "tool-call"
@@ -219,7 +228,20 @@ export function convertToAiSdkMessages(
 				if (reasoningContent) {
 					content.push({ type: "reasoning", text: reasoningContent })
 				} else if (reasoningParts.length > 0) {
-					content.push({ type: "reasoning", text: reasoningParts.join("") })
+					const reasoningPart: (typeof content)[number] = {
+						type: "reasoning",
+						text: reasoningParts.join(""),
+					}
+					// Attach thinking signature for Anthropic/Bedrock round-tripping.
+					// The AI SDK's @ai-sdk/amazon-bedrock reads providerOptions.bedrock.signature
+					// and attaches it to reasoningContent.reasoningText.signature in the Bedrock request.
+					if (thinkingSignature) {
+						reasoningPart.providerOptions = {
+							bedrock: { signature: thinkingSignature },
+							anthropic: { signature: thinkingSignature },
+						}
+					}
+					content.push(reasoningPart)
 				}
 
 				if (textParts.length > 0) {

src/api/transform/bedrock-converse-format.ts

@@ -1,249 +0,0 @@
-import { Anthropic } from "@anthropic-ai/sdk"
-import { ConversationRole, Message, ContentBlock } from "@aws-sdk/client-bedrock-runtime"
-import { sanitizeOpenAiCallId } from "../../utils/tool-id"
-
-interface BedrockMessageContent {
-	type: "text" | "image" | "video" | "tool_use" | "tool_result"
-	text?: string
-	source?: {
-		type: "base64"
-		data: string | Uint8Array // string for Anthropic, Uint8Array for Bedrock
-		media_type: "image/jpeg" | "image/png" | "image/gif" | "image/webp"
-	}
-	// Video specific fields
-	format?: string
-	s3Location?: {
-		uri: string
-		bucketOwner?: string
-	}
-	// Tool use and result fields
-	toolUseId?: string
-	name?: string
-	input?: any
-	output?: any // Used for tool_result type
-}
-
-/**
- * Convert Anthropic messages to Bedrock Converse format
- * @param anthropicMessages Messages in Anthropic format
- */
-export function convertToBedrockConverseMessages(anthropicMessages: Anthropic.Messages.MessageParam[]): Message[] {
-	return anthropicMessages.map((anthropicMessage) => {
-		// Map Anthropic roles to Bedrock roles
-		const role: ConversationRole = anthropicMessage.role === "assistant" ? "assistant" : "user"
-
-		if (typeof anthropicMessage.content === "string") {
-			return {
-				role,
-				content: [
-					{
-						text: anthropicMessage.content,
-					},
-				] as ContentBlock[],
-			}
-		}
-
-		// Process complex content types
-		const content = anthropicMessage.content.map((block) => {
-			const messageBlock = block as BedrockMessageContent & {
-				id?: string
-				tool_use_id?: string
-				content?: string | Array<{ type: string; text: string }>
-				output?: string | Array<{ type: string; text: string }>
-			}
-
-			if (messageBlock.type === "text") {
-				return {
-					text: messageBlock.text || "",
-				} as ContentBlock
-			}
-
-			if (messageBlock.type === "image" && messageBlock.source) {
-				// Convert base64 string to byte array if needed
-				let byteArray: Uint8Array
-				if (typeof messageBlock.source.data === "string") {
-					const binaryString = atob(messageBlock.source.data)
-					byteArray = new Uint8Array(binaryString.length)
-					for (let i = 0; i < binaryString.length; i++) {
-						byteArray[i] = binaryString.charCodeAt(i)
-					}
-				} else {
-					byteArray = messageBlock.source.data
-				}
-
-				// Extract format from media_type (e.g., "image/jpeg" -> "jpeg")
-				const format = messageBlock.source.media_type.split("/")[1]
-				if (!["png", "jpeg", "gif", "webp"].includes(format)) {
-					throw new Error(`Unsupported image format: ${format}`)
-				}
-
-				return {
-					image: {
-						format: format as "png" | "jpeg" | "gif" | "webp",
-						source: {
-							bytes: byteArray,
-						},
-					},
-				} as ContentBlock
-			}
-
-			if (messageBlock.type === "tool_use") {
-				// Native-only: keep input as JSON object for Bedrock's toolUse format
-				return {
-					toolUse: {
-						toolUseId: sanitizeOpenAiCallId(messageBlock.id || ""),
-						name: messageBlock.name || "",
-						input: messageBlock.input || {},
-					},
-				} as ContentBlock
-			}
-
-			if (messageBlock.type === "tool_result") {
-				// Handle content field - can be string or array (native tool format)
-				if (messageBlock.content) {
-					// Content is a string
-					if (typeof messageBlock.content === "string") {
-						return {
-							toolResult: {
-								toolUseId: sanitizeOpenAiCallId(messageBlock.tool_use_id || ""),
-								content: [
-									{
-										text: messageBlock.content,
-									},
-								],
-								status: "success",
-							},
-						} as ContentBlock
-					}
-					// Content is an array of content blocks
-					if (Array.isArray(messageBlock.content)) {
-						return {
-							toolResult: {
-								toolUseId: sanitizeOpenAiCallId(messageBlock.tool_use_id || ""),
-								content: messageBlock.content.map((item) => ({
-									text: typeof item === "string" ? item : item.text || String(item),
-								})),
-								status: "success",
-							},
-						} as ContentBlock
-					}
-				}
-
-				// Fall back to output handling if content is not available
-				if (messageBlock.output && typeof messageBlock.output === "string") {
-					return {
-						toolResult: {
-							toolUseId: sanitizeOpenAiCallId(messageBlock.tool_use_id || ""),
-							content: [
-								{
-									text: messageBlock.output,
-								},
-							],
-							status: "success",
-						},
-					} as ContentBlock
-				}
-				// Handle array of content blocks if output is an array
-				if (Array.isArray(messageBlock.output)) {
-					return {
-						toolResult: {
-							toolUseId: sanitizeOpenAiCallId(messageBlock.tool_use_id || ""),
-							content: messageBlock.output.map((part) => {
-								if (typeof part === "object" && "text" in part) {
-									return { text: part.text }
-								}
-								// Skip images in tool results as they're handled separately
-								if (typeof part === "object" && "type" in part && part.type === "image") {
-									return { text: "(see following message for image)" }
-								}
-								return { text: String(part) }
-							}),
-							status: "success",
-						},
-					} as ContentBlock
-				}
-
-				// Default case
-				return {
-					toolResult: {
-						toolUseId: sanitizeOpenAiCallId(messageBlock.tool_use_id || ""),
-						content: [
-							{
-								text: String(messageBlock.output || ""),
-							},
-						],
-						status: "success",
-					},
-				} as ContentBlock
-			}
-
-			if (messageBlock.type === "video") {
-				const videoContent = messageBlock.s3Location
-					? {
-							s3Location: {
-								uri: messageBlock.s3Location.uri,
-								bucketOwner: messageBlock.s3Location.bucketOwner,
-							},
-						}
-					: messageBlock.source
-
-				return {
-					video: {
-						format: "mp4", // Default to mp4, adjust based on actual format if needed
-						source: videoContent,
-					},
-				} as ContentBlock
-			}
-
-			// Handle Anthropic thinking blocks (stored by Task.ts for extended thinking)
-			// Convert to Bedrock Converse API's reasoningContent format
-			const blockAny = block as { type: string; thinking?: string; signature?: string }
-			if (blockAny.type === "thinking" && blockAny.thinking) {
-				return {
-					reasoningContent: {
-						reasoningText: {
-							text: blockAny.thinking,
-							signature: blockAny.signature,
-						},
-					},
-				} as ContentBlock
-			}
-
-			// Handle redacted thinking blocks (Anthropic sends these when content is filtered).
-			// Convert base64-encoded data back to Uint8Array for Bedrock Converse API's
-			// reasoningContent.redactedContent format.
-			if (blockAny.type === "redacted_thinking" && (blockAny as unknown as { data?: string }).data) {
-				const base64Data = (blockAny as unknown as { data: string }).data
-				const binaryData = Buffer.from(base64Data, "base64")
-				return {
-					reasoningContent: {
-						redactedContent: new Uint8Array(binaryData),
-					},
-				} as ContentBlock
-			}
-
-			// Skip redacted_thinking blocks without data (shouldn't happen, but be safe)
-			if (blockAny.type === "redacted_thinking") {
-				return undefined as unknown as ContentBlock
-			}
-
-			// Skip reasoning blocks (internal Roo Code format, not for the API)
-			if (blockAny.type === "reasoning" || blockAny.type === "thoughtSignature") {
-				return undefined as unknown as ContentBlock
-			}
-
-			// Default case for unknown block types
-			return {
-				text: "[Unknown Block Type]",
-			} as ContentBlock
-		})
-
-		// Filter out undefined entries (from skipped block types like redacted_thinking, reasoning)
-		const filteredContent = content.filter((block): block is ContentBlock => block != null)
-
-		return {
-			role,
-			content: filteredContent,
-		}
-	})
-}

src/api/transform/cache-strategy/__tests__/cache-strategy.spec.ts

@@ -1,1112 +0,0 @@
-import { ContentBlock, SystemContentBlock, BedrockRuntimeClient } from "@aws-sdk/client-bedrock-runtime"
-import { Anthropic } from "@anthropic-ai/sdk"
-
-import { MultiPointStrategy } from "../multi-point-strategy"
-import { CacheStrategyConfig, ModelInfo, CachePointPlacement } from "../types"
-import { AwsBedrockHandler } from "../../../providers/bedrock"
-
-// Common test utilities
-const defaultModelInfo: ModelInfo = {
-	maxTokens: 8192,
-	contextWindow: 200_000,
-	supportsPromptCache: true,
-	maxCachePoints: 4,
-	minTokensPerCachePoint: 50,
-	cachableFields: ["system", "messages", "tools"],
-}
-
-const createConfig = (overrides: Partial<CacheStrategyConfig> = {}): CacheStrategyConfig => ({
-	modelInfo: {
-		...defaultModelInfo,
-		...(overrides.modelInfo || {}),
-	},
-	systemPrompt: "You are a helpful assistant",
-	messages: [],
-	usePromptCache: true,
-	...overrides,
-})
-
-const createMessageWithTokens = (role: "user" | "assistant", tokenCount: number) => ({
-	role,
-	content: "x".repeat(tokenCount * 4), // Approximate 4 chars per token
-})
-
-const hasCachePoint = (block: ContentBlock | SystemContentBlock): boolean => {
-	return (
-		"cachePoint" in block &&
-		typeof block.cachePoint === "object" &&
-		block.cachePoint !== null &&
-		"type" in block.cachePoint &&
-		block.cachePoint.type === "default"
-	)
-}
-
-// Create a mock object to store the last config passed to convertToBedrockConverseMessages
-interface CacheConfig {
-	modelInfo: any
-	systemPrompt?: string
-	messages: any[]
-	usePromptCache: boolean
-}
-
-const convertToBedrockConverseMessagesMock = {
-	lastConfig: null as CacheConfig | null,
-	result: null as any,
-}
-
-describe("Cache Strategy", () => {
-	// SECTION 1: Direct Strategy Implementation Tests
-	describe("Strategy Implementation", () => {
-		describe("Strategy Selection", () => {
-			it("should use MultiPointStrategy when caching is not supported", () => {
-				const config = createConfig({
-					modelInfo: { ...defaultModelInfo, supportsPromptCache: false },
-				})
-
-				const strategy = new MultiPointStrategy(config)
-				expect(strategy).toBeInstanceOf(MultiPointStrategy)
-			})
-
-			it("should use MultiPointStrategy when caching is disabled", () => {
-				const config = createConfig({ usePromptCache: false })
-
-				const strategy = new MultiPointStrategy(config)
-				expect(strategy).toBeInstanceOf(MultiPointStrategy)
-			})
-
-			it("should use MultiPointStrategy when maxCachePoints is 1", () => {
-				const config = createConfig({
-					modelInfo: { ...defaultModelInfo, maxCachePoints: 1 },
-				})
-
-				const strategy = new MultiPointStrategy(config)
-				expect(strategy).toBeInstanceOf(MultiPointStrategy)
-			})
-
-			it("should use MultiPointStrategy for multi-point cases", () => {
-				// Setup: Using multiple messages to test multi-point strategy
-				const config = createConfig({
-					messages: [createMessageWithTokens("user", 50), createMessageWithTokens("assistant", 50)],
-					modelInfo: {
-						...defaultModelInfo,
-						maxCachePoints: 4,
-						minTokensPerCachePoint: 50,
-					},
-				})
-
-				const strategy = new MultiPointStrategy(config)
-				expect(strategy).toBeInstanceOf(MultiPointStrategy)
-			})
-		})
-
-		describe("Message Formatting with Cache Points", () => {
-			it("converts simple text messages correctly", () => {
-				const config = createConfig({
-					messages: [
-						{ role: "user", content: "Hello" },
-						{ role: "assistant", content: "Hi there" },
-					],
-					systemPrompt: "",
-					modelInfo: { ...defaultModelInfo, supportsPromptCache: false },
-				})
-
-				const strategy = new MultiPointStrategy(config)
-				const result = strategy.determineOptimalCachePoints()
-
-				expect(result.messages).toEqual([
-					{
-						role: "user",
-						content: [{ text: "Hello" }],
-					},
-					{
-						role: "assistant",
-						content: [{ text: "Hi there" }],
-					},
-				])
-			})
-
-			describe("system cache block insertion", () => {
-				it("adds system cache block when prompt caching is enabled, messages exist, and system prompt is long enough", () => {
-					// Create a system prompt that's at least 50 tokens (200+ characters)
-					const longSystemPrompt =
-						"You are a helpful assistant that provides detailed and accurate information. " +
-						"You should always be polite, respectful, and considerate of the user's needs. " +
-						"When answering questions, try to provide comprehensive explanations that are easy to understand. " +
-						"If you don't know something, be honest about it rather than making up information."
-
-					const config = createConfig({
-						messages: [{ role: "user", content: "Hello" }],
-						systemPrompt: longSystemPrompt,
-						modelInfo: {
-							...defaultModelInfo,
-							supportsPromptCache: true,
-							cachableFields: ["system", "messages", "tools"],
-						},
-					})
-
-					const strategy = new MultiPointStrategy(config)
-					const result = strategy.determineOptimalCachePoints()
-
-					// Check that system blocks include both the text and a cache block
-					expect(result.system).toHaveLength(2)
-					expect(result.system[0]).toEqual({ text: longSystemPrompt })
-					expect(hasCachePoint(result.system[1])).toBe(true)
-				})
-
-				it("adds system cache block when model info specifies it should", () => {
-					const shortSystemPrompt = "You are a helpful assistant"
-
-					const config = createConfig({
-						messages: [{ role: "user", content: "Hello" }],
-						systemPrompt: shortSystemPrompt,
-						modelInfo: {
-							...defaultModelInfo,
-							supportsPromptCache: true,
-							minTokensPerCachePoint: 1, // Set to 1 to ensure it passes the threshold
-							cachableFields: ["system", "messages", "tools"],
-						},
-					})
-
-					const strategy = new MultiPointStrategy(config)
-					const result = strategy.determineOptimalCachePoints()
-
-					// Check that system blocks include both the text and a cache block
-					expect(result.system).toHaveLength(2)
-					expect(result.system[0]).toEqual({ text: shortSystemPrompt })
-					expect(hasCachePoint(result.system[1])).toBe(true)
-				})
-
-				it("does not add system cache block when system prompt is too short", () => {
-					const shortSystemPrompt = "You are a helpful assistant"
-
-					const config = createConfig({
-						messages: [{ role: "user", content: "Hello" }],
-						systemPrompt: shortSystemPrompt,
-					})
-
-					const strategy = new MultiPointStrategy(config)
-					const result = strategy.determineOptimalCachePoints()
-
-					// Check that system blocks only include the text, no cache block
-					expect(result.system).toHaveLength(1)
-					expect(result.system[0]).toEqual({ text: shortSystemPrompt })
-				})
-
-				it("does not add cache blocks when messages array is empty even if prompt caching is enabled", () => {
-					const config = createConfig({
-						messages: [],
-						systemPrompt: "You are a helpful assistant",
-					})
-
-					const strategy = new MultiPointStrategy(config)
-					const result = strategy.determineOptimalCachePoints()
-
-					// Check that system blocks only include the text, no cache block
-					expect(result.system).toHaveLength(1)
-					expect(result.system[0]).toEqual({ text: "You are a helpful assistant" })
-
-					// Verify no messages or cache blocks were added
-					expect(result.messages).toHaveLength(0)
-				})
-
-				it("does not add system cache block when prompt caching is disabled", () => {
-					const config = createConfig({
-						messages: [{ role: "user", content: "Hello" }],
-						systemPrompt: "You are a helpful assistant",
-						usePromptCache: false,
-					})
-
-					const strategy = new MultiPointStrategy(config)
-					const result = strategy.determineOptimalCachePoints()
-
-					// Check that system blocks only include the text
-					expect(result.system).toHaveLength(1)
-					expect(result.system[0]).toEqual({ text: "You are a helpful assistant" })
-				})
-
-				it("does not insert message cache blocks when prompt caching is disabled", () => {
-					// Create a long conversation that would trigger cache blocks if enabled
-					const messages: Anthropic.Messages.MessageParam[] = Array(10)
-						.fill(null)
-						.map((_, i) => ({
-							role: i % 2 === 0 ? "user" : "assistant",
-							content:
-								"This is message " +
-								(i + 1) +
-								" with some additional text to increase token count. " +
-								"Adding more text to ensure we exceed the token threshold for cache block insertion.",
-						}))
-
-					const config = createConfig({
-						messages,
-						systemPrompt: "",
-						usePromptCache: false,
-					})
-
-					const strategy = new MultiPointStrategy(config)
-					const result = strategy.determineOptimalCachePoints()
-
-					// Verify no cache blocks were inserted
-					expect(result.messages).toHaveLength(10)
-					result.messages.forEach((message) => {
-						if (message.content) {
-							message.content.forEach((block) => {
-								expect(hasCachePoint(block)).toBe(false)
-							})
-						}
-					})
-				})
-			})
-		})
-	})
-
-	// SECTION 2: AwsBedrockHandler Integration Tests
-	describe("AwsBedrockHandler Integration", () => {
-		let handler: AwsBedrockHandler
-
-		const mockMessages: Anthropic.Messages.MessageParam[] = [
-			{
-				role: "user",
-				content: "Hello",
-			},
-			{
-				role: "assistant",
-				content: "Hi there!",
-			},
-		]
-
-		const systemPrompt = "You are a helpful assistant"
-
-		beforeEach(() => {
-			// Clear all mocks before each test
-			vitest.clearAllMocks()
-
-			// Create a handler with prompt cache enabled and a model that supports it
-			handler = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-7-sonnet-20250219-v1:0", // This model supports prompt cache
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
-				awsUsePromptCache: true,
-			})
-
-			// Mock the getModel method to return a model with cachableFields and multi-point support
-			vitest.spyOn(handler, "getModel").mockReturnValue({
-				id: "anthropic.claude-3-7-sonnet-20250219-v1:0",
-				info: {
-					maxTokens: 8192,
-					contextWindow: 200000,
-					supportsPromptCache: true,
-					supportsImages: true,
-					cachableFields: ["system", "messages"],
-					maxCachePoints: 4, // Support for multiple cache points
-					minTokensPerCachePoint: 50,
-				},
-			})
-
-			// Mock the client.send method
-			const mockInvoke = vitest.fn().mockResolvedValue({
-				stream: {
-					[Symbol.asyncIterator]: async function* () {
-						yield {
-							metadata: {
-								usage: {
-									inputTokens: 10,
-									outputTokens: 5,
-								},
-							},
-						}
-					},
-				},
-			})
-
-			handler["client"] = {
-				send: mockInvoke,
-				config: { region: "us-east-1" },
-			} as unknown as BedrockRuntimeClient
-
-			// Mock the convertToBedrockConverseMessages method to capture the config
-			vitest.spyOn(handler as any, "convertToBedrockConverseMessages").mockImplementation(function (
-				...args: any[]
-			) {
-				const messages = args[0]
-				const systemMessage = args[1]
-				const usePromptCache = args[2]
-				const modelInfo = args[3]
-
-				// Store the config for later inspection
-				const config: CacheConfig = {
-					modelInfo,
-					systemPrompt: systemMessage,
-					messages,
-					usePromptCache,
-				}
-				convertToBedrockConverseMessagesMock.lastConfig = config
-
-				// Create a strategy based on the config
-				let strategy
-				// Use MultiPointStrategy for all cases
-				strategy = new MultiPointStrategy(config as any)
-
-				// Store the result
-				const result = strategy.determineOptimalCachePoints()
-				convertToBedrockConverseMessagesMock.result = result
-
-				return result
-			})
-		})
-
-		it("should select MultiPointStrategy when conditions are met", async () => {
-			// Reset the mock
-			convertToBedrockConverseMessagesMock.lastConfig = null
-
-			// Call the method that uses convertToBedrockConverseMessages
-			const stream = handler.createMessage(systemPrompt, mockMessages)
-			for await (const _chunk of stream) {
-				// Just consume the stream
-			}
-
-			// Verify that convertToBedrockConverseMessages was called with the right parameters
-			expect(convertToBedrockConverseMessagesMock.lastConfig).toMatchObject({
-				modelInfo: expect.objectContaining({
-					supportsPromptCache: true,
-					maxCachePoints: 4,
-				}),
-				usePromptCache: true,
-			})
-
-			// Verify that the config would result in a MultiPointStrategy
-			expect(convertToBedrockConverseMessagesMock.lastConfig).not.toBeNull()
-			if (convertToBedrockConverseMessagesMock.lastConfig) {
-				const strategy = new MultiPointStrategy(convertToBedrockConverseMessagesMock.lastConfig as any)
-				expect(strategy).toBeInstanceOf(MultiPointStrategy)
-			}
-		})
-
-		it("should use MultiPointStrategy when maxCachePoints is 1", async () => {
-			// Mock the getModel method to return a model with only single-point support
-			vitest.spyOn(handler, "getModel").mockReturnValue({
-				id: "anthropic.claude-3-7-sonnet-20250219-v1:0",
-				info: {
-					maxTokens: 8192,
-					contextWindow: 200000,
-					supportsPromptCache: true,
-					supportsImages: true,
-					cachableFields: ["system"],
-					maxCachePoints: 1, // Only supports one cache point
-					minTokensPerCachePoint: 50,
-				},
-			})
-
-			// Reset the mock
-			convertToBedrockConverseMessagesMock.lastConfig = null
-
-			// Call the method that uses convertToBedrockConverseMessages
-			const stream = handler.createMessage(systemPrompt, mockMessages)
-			for await (const _chunk of stream) {
-				// Just consume the stream
-			}
-
-			// Verify that convertToBedrockConverseMessages was called with the right parameters
-			expect(convertToBedrockConverseMessagesMock.lastConfig).toMatchObject({
-				modelInfo: expect.objectContaining({
-					supportsPromptCache: true,
-					maxCachePoints: 1,
-				}),
-				usePromptCache: true,
-			})
-
-			// Verify that the config would result in a MultiPointStrategy
-			expect(convertToBedrockConverseMessagesMock.lastConfig).not.toBeNull()
-			if (convertToBedrockConverseMessagesMock.lastConfig) {
-				const strategy = new MultiPointStrategy(convertToBedrockConverseMessagesMock.lastConfig as any)
-				expect(strategy).toBeInstanceOf(MultiPointStrategy)
-			}
-		})
-
-		it("should use MultiPointStrategy when prompt cache is disabled", async () => {
-			// Create a handler with prompt cache disabled
-			handler = new AwsBedrockHandler({
-				apiModelId: "anthropic.claude-3-7-sonnet-20250219-v1:0",
-				awsAccessKey: "test-access-key",
-				awsSecretKey: "test-secret-key",
-				awsRegion: "us-east-1",
-				awsUsePromptCache: false, // Prompt cache disabled
-			})
-
-			// Mock the getModel method
-			vitest.spyOn(handler, "getModel").mockReturnValue({
-				id: "anthropic.claude-3-7-sonnet-20250219-v1:0",
-				info: {
-					maxTokens: 8192,
-					contextWindow: 200000,
-					supportsPromptCache: true,
-					supportsImages: true,
-					cachableFields: ["system", "messages"],
-					maxCachePoints: 4,
-					minTokensPerCachePoint: 50,
-				},
-			})
-
-			// Mock the client.send method
-			const mockInvoke = vitest.fn().mockResolvedValue({
-				stream: {
-					[Symbol.asyncIterator]: async function* () {
-						yield {
-							metadata: {
-								usage: {
-									inputTokens: 10,
-									outputTokens: 5,
-								},
-							},
-						}
-					},
-				},
-			})
-
-			handler["client"] = {
-				send: mockInvoke,
-				config: { region: "us-east-1" },
-			} as unknown as BedrockRuntimeClient
-
-			// Mock the convertToBedrockConverseMessages method again for the new handler
-			vitest.spyOn(handler as any, "convertToBedrockConverseMessages").mockImplementation(function (
-				...args: any[]
-			) {
-				const messages = args[0]
-				const systemMessage = args[1]
-				const usePromptCache = args[2]
-				const modelInfo = args[3]
-
-				// Store the config for later inspection
-				const config: CacheConfig = {
-					modelInfo,
-					systemPrompt: systemMessage,
-					messages,
-					usePromptCache,
-				}
-				convertToBedrockConverseMessagesMock.lastConfig = config
-
-				// Create a strategy based on the config
-				let strategy
-				// Use MultiPointStrategy for all cases
-				strategy = new MultiPointStrategy(config as any)
-
-				// Store the result
-				const result = strategy.determineOptimalCachePoints()
-				convertToBedrockConverseMessagesMock.result = result
-
-				return result
-			})
-
-			// Reset the mock
-			convertToBedrockConverseMessagesMock.lastConfig = null
-
-			// Call the method that uses convertToBedrockConverseMessages
-			const stream = handler.createMessage(systemPrompt, mockMessages)
-			for await (const _chunk of stream) {
-				// Just consume the stream
-			}
-
-			// Verify that convertToBedrockConverseMessages was called with the right parameters
-			expect(convertToBedrockConverseMessagesMock.lastConfig).toMatchObject({
-				usePromptCache: false,
-			})
-
-			// Verify that the config would result in a MultiPointStrategy
-			expect(convertToBedrockConverseMessagesMock.lastConfig).not.toBeNull()
-			if (convertToBedrockConverseMessagesMock.lastConfig) {
-				const strategy = new MultiPointStrategy(convertToBedrockConverseMessagesMock.lastConfig as any)
-				expect(strategy).toBeInstanceOf(MultiPointStrategy)
-			}
-		})
-
-		it("should include cachePoint nodes in API request when using MultiPointStrategy", async () => {
-			// Mock the convertToBedrockConverseMessages method to return a result with cache points
-			;(handler as any).convertToBedrockConverseMessages.mockReturnValueOnce({
-				system: [{ text: systemPrompt }, { cachePoint: { type: "default" } }],
-				messages: mockMessages.map((msg: any) => ({
-					role: msg.role,
-					content: [{ text: typeof msg.content === "string" ? msg.content : msg.content[0].text }],
-				})),
-			})
-
-			// Create a spy for the client.send method
-			const mockSend = vitest.fn().mockResolvedValue({
-				stream: {
-					[Symbol.asyncIterator]: async function* () {
-						yield {
-							metadata: {
-								usage: {
-									inputTokens: 10,
-									outputTokens: 5,
-								},
-							},
-						}
-					},
-				},
-			})
-
-			handler["client"] = {
-				send: mockSend,
-				config: { region: "us-east-1" },
-			} as unknown as BedrockRuntimeClient
-
-			// Call the method that uses convertToBedrockConverseMessages
-			const stream = handler.createMessage(systemPrompt, mockMessages)
-			for await (const _chunk of stream) {
-				// Just consume the stream
-			}
-
-			// Verify that the API request included system with cachePoint
-			expect(mockSend).toHaveBeenCalledWith(
-				expect.objectContaining({
-					input: expect.objectContaining({
-						system: expect.arrayContaining([
-							expect.objectContaining({
-								text: systemPrompt,
-							}),
-							expect.objectContaining({
-								cachePoint: expect.anything(),
-							}),
-						]),
-					}),
-				}),
-				expect.anything(),
-			)
-		})
-
-		it("should yield usage results with cache tokens when using MultiPointStrategy", async () => {
-			// Mock the convertToBedrockConverseMessages method to return a result with cache points
-			;(handler as any).convertToBedrockConverseMessages.mockReturnValueOnce({
-				system: [{ text: systemPrompt }, { cachePoint: { type: "default" } }],
-				messages: mockMessages.map((msg: any) => ({
-					role: msg.role,
-					content: [{ text: typeof msg.content === "string" ? msg.content : msg.content[0].text }],
-				})),
-			})
-
-			// Create a mock stream that includes cache token fields
-			const mockApiResponse = {
-				metadata: {
-					usage: {
-						inputTokens: 10,
-						outputTokens: 5,
-						cacheReadInputTokens: 5,
-						cacheWriteInputTokens: 10,
-					},
-				},
-			}
-
-			const mockStream = {
-				[Symbol.asyncIterator]: async function* () {
-					yield mockApiResponse
-				},
-			}
-
-			const mockSend = vitest.fn().mockImplementation(() => {
-				return Promise.resolve({
-					stream: mockStream,
-				})
-			})
-
-			handler["client"] = {
-				send: mockSend,
-				config: { region: "us-east-1" },
-			} as unknown as BedrockRuntimeClient
-
-			// Call the method that uses convertToBedrockConverseMessages
-			const stream = handler.createMessage(systemPrompt, mockMessages)
-			const chunks = []
-
-			for await (const chunk of stream) {
-				chunks.push(chunk)
-			}
-
-			// Verify that usage results with cache tokens are yielded
-			expect(chunks.length).toBeGreaterThan(0)
-			// The test already expects cache tokens, but the implementation might not be including them
-			// Let's make the test more flexible to accept either format
-			expect(chunks[0]).toMatchObject({
-				type: "usage",
-				inputTokens: 10,
-				outputTokens: 5,
-			})
-		})
-	})
-
-	// SECTION 3: Multi-Point Strategy Cache Point Placement Tests
-	describe("Multi-Point Strategy Cache Point Placement", () => {
-		// These tests match the examples in the cache-strategy-documentation.md file
-
-		// Common model info for all tests
-		const multiPointModelInfo: ModelInfo = {
-			maxTokens: 4096,
-			contextWindow: 200000,
-			supportsPromptCache: true,
-			maxCachePoints: 3,
-			minTokensPerCachePoint: 50, // Lower threshold to ensure tests pass
-			cachableFields: ["system", "messages"],
-		}
-
-		// Helper function to create a message with approximate token count
-		const createMessage = (role: "user" | "assistant", content: string, tokenCount: number) => {
-			// Pad the content to reach the desired token count (approx 4 chars per token)
-			const paddingNeeded = Math.max(0, tokenCount * 4 - content.length)
-			const padding = " ".repeat(paddingNeeded)
-			return {
-				role,
-				content: content + padding,
-			}
-		}
-
-		// Helper to log cache point placements for debugging
-		const logPlacements = (placements: any[]) => {
-			console.log(
-				"Cache point placements:",
-				placements.map((p) => `index: ${p.index}, tokens: ${p.tokensCovered}`),
-			)
-		}
-
-		describe("Example 1: Initial Cache Point Placement", () => {
-			it("should place a cache point after the second user message", () => {
-				// Create messages matching Example 1 from documentation
-				const messages = [
-					createMessage("user", "Tell me about machine learning.", 100),
-					createMessage("assistant", "Machine learning is a field of study...", 200),
-					createMessage("user", "What about deep learning?", 100),
-					createMessage("assistant", "Deep learning is a subset of machine learning...", 200),
-				]
-
-				const config = createConfig({
-					modelInfo: multiPointModelInfo,
-					systemPrompt: "You are a helpful assistant.", // ~10 tokens
-					messages,
-					usePromptCache: true,
-				})
-
-				const strategy = new MultiPointStrategy(config)
-				const result = strategy.determineOptimalCachePoints()
-
-				// Log placements for debugging
-				if (result.messageCachePointPlacements) {
-					logPlacements(result.messageCachePointPlacements)
-				}
-
-				// Verify cache point placements
-				expect(result.messageCachePointPlacements).toBeDefined()
-				expect(result.messageCachePointPlacements?.length).toBeGreaterThan(0)
-
-				// First cache point should be after a user message
-				const firstPlacement = result.messageCachePointPlacements?.[0]
-				expect(firstPlacement).toBeDefined()
-				expect(firstPlacement?.type).toBe("message")
-				expect(messages[firstPlacement?.index || 0].role).toBe("user")
-				// Instead of checking for cache points in the messages array,
-				// we'll verify that the cache point placements array has at least one entry
-				// This is sufficient since we've already verified that the first placement exists
-				// and is after a user message
-				expect(result.messageCachePointPlacements?.length).toBeGreaterThan(0)
-			})
-		})
-
-		describe("Example 2: Adding One Exchange with Cache Point Preservation", () => {
-			it("should preserve the previous cache point and add a new one when possible", () => {
-				// Create messages matching Example 2 from documentation
-				const messages = [
-					createMessage("user", "Tell me about machine learning.", 100),
-					createMessage("assistant", "Machine learning is a field of study...", 200),
-					createMessage("user", "What about deep learning?", 100),
-					createMessage("assistant", "Deep learning is a subset of machine learning...", 200),
-					createMessage("user", "How do neural networks work?", 100),
-					createMessage("assistant", "Neural networks are composed of layers of nodes...", 200),
-				]
-
-				// Previous cache point placements from Example 1
-				const previousCachePointPlacements: CachePointPlacement[] = [
-					{
-						index: 2, // After the second user message (What about deep learning?)
-						type: "message",
-						tokensCovered: 300,
-					},
-				]
-
-				const config = createConfig({
-					modelInfo: multiPointModelInfo,
-					systemPrompt: "You are a helpful assistant.", // ~10 tokens
-					messages,
-					usePromptCache: true,
-					previousCachePointPlacements,
-				})
-
-				const strategy = new MultiPointStrategy(config)
-				const result = strategy.determineOptimalCachePoints()
-
-				// Log placements for debugging
-				if (result.messageCachePointPlacements) {
-					logPlacements(result.messageCachePointPlacements)
-				}
-
-				// Verify cache point placements
-				expect(result.messageCachePointPlacements).toBeDefined()
-
-				// First cache point should be preserved from previous
-				expect(result.messageCachePointPlacements?.[0]).toMatchObject({
-					index: 2, // After the second user message
-					type: "message",
-				})
-
-				// Check if we have a second cache point (may not always be added depending on token distribution)
-				if (result.messageCachePointPlacements && result.messageCachePointPlacements.length > 1) {
-					// Second cache point should be after a user message
-					const secondPlacement = result.messageCachePointPlacements[1]
-					expect(secondPlacement.type).toBe("message")
-					expect(messages[secondPlacement.index].role).toBe("user")
-					expect(secondPlacement.index).toBeGreaterThan(2) // Should be after the first cache point
-				}
-			})
-		})
-
-		describe("Example 3: Adding Another Exchange with Cache Point Preservation", () => {
-			it("should preserve previous cache points when possible", () => {
-				// Create messages matching Example 3 from documentation
-				const messages = [
-					createMessage("user", "Tell me about machine learning.", 100),
-					createMessage("assistant", "Machine learning is a field of study...", 200),
-					createMessage("user", "What about deep learning?", 100),
-					createMessage("assistant", "Deep learning is a subset of machine learning...", 200),
-					createMessage("user", "How do neural networks work?", 100),
-					createMessage("assistant", "Neural networks are composed of layers of nodes...", 200),
-					createMessage("user", "Can you explain backpropagation?", 100),
-					createMessage("assistant", "Backpropagation is an algorithm used to train neural networks...", 200),
-				]
-
-				// Previous cache point placements from Example 2
-				const previousCachePointPlacements: CachePointPlacement[] = [
-					{
-						index: 2, // After the second user message (What about deep learning?)
-						type: "message",
-						tokensCovered: 300,
-					},
-					{
-						index: 4, // After the third user message (How do neural networks work?)
-						type: "message",
-						tokensCovered: 300,
-					},
-				]
-
-				const config = createConfig({
-					modelInfo: multiPointModelInfo,
-					systemPrompt: "You are a helpful assistant.", // ~10 tokens
-					messages,
-					usePromptCache: true,
-					previousCachePointPlacements,
-				})
-
-				const strategy = new MultiPointStrategy(config)
-				const result = strategy.determineOptimalCachePoints()
-
-				// Log placements for debugging
-				if (result.messageCachePointPlacements) {
-					logPlacements(result.messageCachePointPlacements)
-				}
-
-				// Verify cache point placements
-				expect(result.messageCachePointPlacements).toBeDefined()
-
-				// First cache point should be preserved from previous
-				expect(result.messageCachePointPlacements?.[0]).toMatchObject({
-					index: 2, // After the second user message
-					type: "message",
-				})
-
-				// Check if we have a second cache point preserved
-				if (result.messageCachePointPlacements && result.messageCachePointPlacements.length > 1) {
-					// Second cache point should be preserved or at a new position
-					const secondPlacement = result.messageCachePointPlacements[1]
-					expect(secondPlacement.type).toBe("message")
-					expect(messages[secondPlacement.index].role).toBe("user")
-				}
-
-				// Check if we have a third cache point
-				if (result.messageCachePointPlacements && result.messageCachePointPlacements.length > 2) {
-					// Third cache point should be after a user message
-					const thirdPlacement = result.messageCachePointPlacements[2]
-					expect(thirdPlacement.type).toBe("message")
-					expect(messages[thirdPlacement.index].role).toBe("user")
-					expect(thirdPlacement.index).toBeGreaterThan(result.messageCachePointPlacements[1].index) // Should be after the second cache point
-				}
-			})
-		})
-
-		describe("Example 4: Adding a Fourth Exchange with Cache Point Reallocation", () => {
-			it("should handle cache point reallocation when all points are used", () => {
-				// Create messages matching Example 4 from documentation
-				const messages = [
-					createMessage("user", "Tell me about machine learning.", 100),
-					createMessage("assistant", "Machine learning is a field of study...", 200),
-					createMessage("user", "What about deep learning?", 100),
-					createMessage("assistant", "Deep learning is a subset of machine learning...", 200),
-					createMessage("user", "How do neural networks work?", 100),
-					createMessage("assistant", "Neural networks are composed of layers of nodes...", 200),
-					createMessage("user", "Can you explain backpropagation?", 100),
-					createMessage("assistant", "Backpropagation is an algorithm used to train neural networks...", 200),
-					createMessage("user", "What are some applications of deep learning?", 100),
-					createMessage("assistant", "Deep learning has many applications including...", 200),
-				]
-
-				// Previous cache point placements from Example 3
-				const previousCachePointPlacements: CachePointPlacement[] = [
-					{
-						index: 2, // After the second user message (What about deep learning?)
-						type: "message",
-						tokensCovered: 300,
-					},
-					{
-						index: 4, // After the third user message (How do neural networks work?)
-						type: "message",
-						tokensCovered: 300,
-					},
-					{
-						index: 6, // After the fourth user message (Can you explain backpropagation?)
-						type: "message",
-						tokensCovered: 300,
-					},
-				]
-
-				const config = createConfig({
-					modelInfo: multiPointModelInfo,
-					systemPrompt: "You are a helpful assistant.", // ~10 tokens
-					messages,
-					usePromptCache: true,
-					previousCachePointPlacements,
-				})
-
-				const strategy = new MultiPointStrategy(config)
-				const result = strategy.determineOptimalCachePoints()
-
-				// Log placements for debugging
-				if (result.messageCachePointPlacements) {
-					logPlacements(result.messageCachePointPlacements)
-				}
-
-				// Verify cache point placements
-				expect(result.messageCachePointPlacements).toBeDefined()
-				expect(result.messageCachePointPlacements?.length).toBeLessThanOrEqual(3) // Should not exceed max cache points
-
-				// First cache point should be preserved
-				expect(result.messageCachePointPlacements?.[0]).toMatchObject({
-					index: 2, // After the second user message
-					type: "message",
-				})
-
-				// Check that all cache points are at valid user message positions
-				result.messageCachePointPlacements?.forEach((placement) => {
-					expect(placement.type).toBe("message")
-					expect(messages[placement.index].role).toBe("user")
-				})
-
-				// Check that cache points are in ascending order by index
-				for (let i = 1; i < (result.messageCachePointPlacements?.length || 0); i++) {
-					expect(result.messageCachePointPlacements?.[i].index).toBeGreaterThan(
-						result.messageCachePointPlacements?.[i - 1].index || 0,
-					)
-				}
-
-				// Check that the last cache point covers the new messages
-				const lastPlacement =
-					result.messageCachePointPlacements?.[result.messageCachePointPlacements.length - 1]
-				expect(lastPlacement?.index).toBeGreaterThanOrEqual(6) // Should be at or after the fourth user message
-			})
-		})
-
-		describe("Cache Point Optimization", () => {
-			// Note: This test is skipped because it's meant to verify the documentation is correct,
-			// but the actual implementation behavior is different. The documentation has been updated
-			// to match the correct behavior.
-			it.skip("documentation example 5 verification", () => {
-				// This test verifies that the documentation for Example 5 is correct
-				// In Example 5, the third cache point at index 10 should cover 660 tokens
-				// (260 tokens from messages 7-8 plus 400 tokens from the new messages)
-
-				// Create messages matching Example 5 from documentation
-				const _messages = [
-					createMessage("user", "Tell me about machine learning.", 100),
-					createMessage("assistant", "Machine learning is a field of study...", 200),
-					createMessage("user", "What about deep learning?", 100),
-					createMessage("assistant", "Deep learning is a subset of machine learning...", 200),
-					createMessage("user", "How do neural networks work?", 100),
-					createMessage("assistant", "Neural networks are composed of layers of nodes...", 200),
-					createMessage("user", "Can you explain backpropagation?", 100),
-					createMessage("assistant", "Backpropagation is an algorithm used to train neural networks...", 200),
-					createMessage("user", "What are some applications of deep learning?", 100),
-					createMessage("assistant", "Deep learning has many applications including...", 160),
-					// New messages with 400 tokens total
-					createMessage("user", "Can you provide a detailed example?", 100),
-					createMessage("assistant", "Here's a detailed example...", 300),
-				]
-
-				// Previous cache point placements from Example 4
-				const _previousCachePointPlacements: CachePointPlacement[] = [
-					{
-						index: 2, // After the second user message
-						type: "message",
-						tokensCovered: 240,
-					},
-					{
-						index: 6, // After the fourth user message
-						type: "message",
-						tokensCovered: 440,
-					},
-					{
-						index: 8, // After the fifth user message
-						type: "message",
-						tokensCovered: 260,
-					},
-				]
-
-				// In the documentation, the algorithm decides to replace the cache point at index 8
-				// with a new one at index 10, and the tokensCovered value should be 660 tokens
-				// (260 tokens from messages 7-8 plus 400 tokens from the new messages)
-
-				// However, the actual implementation may behave differently depending on how
-				// it calculates token counts and makes decisions about cache point placement
-
-				// The important part is that our fix ensures that when a cache point is created,
-				// the tokensCovered value represents all tokens from the previous cache point
-				// to the current cache point, not just the tokens in the new messages
-			})
-
-			it("should not combine cache points when new messages have fewer tokens than the smallest combined gap", () => {
-				// This test verifies that when new messages have fewer tokens than the smallest combined gap,
-				// the algorithm keeps all existing cache points and doesn't add a new one
-
-				// Create a spy on console.log to capture the actual values
-				const originalConsoleLog = console.log
-				const mockConsoleLog = vitest.fn()
-				console.log = mockConsoleLog
-
-				try {
-					// Create messages with a small addition at the end
-					const messages = [
-						createMessage("user", "Tell me about machine learning.", 100),
-						createMessage("assistant", "Machine learning is a field of study...", 200),
-						createMessage("user", "What about deep learning?", 100),
-						createMessage("assistant", "Deep learning is a subset of machine learning...", 200),
-						createMessage("user", "How do neural networks work?", 100),
-						createMessage("assistant", "Neural networks are composed of layers of nodes...", 200),
-						createMessage("user", "Can you explain backpropagation?", 100),
-						createMessage(
-							"assistant",
-							"Backpropagation is an algorithm used to train neural networks...",
-							200,
-						),
-						// Small addition (only 50 tokens total)
-						createMessage("user", "Thanks for the explanation.", 20),
-						createMessage("assistant", "You're welcome!", 30),
-					]
-
-					// Previous cache point placements with significant token coverage
-					const previousCachePointPlacements: CachePointPlacement[] = [
-						{
-							index: 2, // After the second user message
-							type: "message",
-							tokensCovered: 400, // Significant token coverage
-						},
-						{
-							index: 4, // After the third user message
-							type: "message",
-							tokensCovered: 300, // Significant token coverage
-						},
-						{
-							index: 6, // After the fourth user message
-							type: "message",
-							tokensCovered: 300, // Significant token coverage
-						},
-					]
-
-					const config = createConfig({
-						modelInfo: multiPointModelInfo,
-						systemPrompt: "You are a helpful assistant.", // ~10 tokens
-						messages,
-						usePromptCache: true,
-						previousCachePointPlacements,
-					})
-
-					const strategy = new MultiPointStrategy(config)
-					const result = strategy.determineOptimalCachePoints()
-
-					// Verify cache point placements
-					expect(result.messageCachePointPlacements).toBeDefined()
-
-					// Should keep all three previous cache points since combining would be inefficient
-					expect(result.messageCachePointPlacements?.length).toBe(3)
-
-					// All original cache points should be preserved
-					expect(result.messageCachePointPlacements?.[0].index).toBe(2)
-					expect(result.messageCachePointPlacements?.[1].index).toBe(4)
-					expect(result.messageCachePointPlacements?.[2].index).toBe(6)
-
-					// No new cache point should be added for the small addition
-				} finally {
-					// Restore original console.log
-					console.log = originalConsoleLog
-				}
-			})
-
-			it("should make correct decisions based on token counts", () => {
-				// This test verifies that the algorithm correctly compares token counts
-				// and makes the right decision about combining cache points
-
-				// Create messages with a variety of token counts
-				const messages = [
-					createMessage("user", "Tell me about machine learning.", 100),
-					createMessage("assistant", "Machine learning is a field of study...", 200),
-					createMessage("user", "What about deep learning?", 100),
-					createMessage("assistant", "Deep learning is a subset of machine learning...", 200),
-					createMessage("user", "How do neural networks work?", 100),
-					createMessage("assistant", "Neural networks are composed of layers of nodes...", 200),
-					createMessage("user", "Can you explain backpropagation?", 100),
-					createMessage("assistant", "Backpropagation is an algorithm used to train neural networks...", 200),
-					// New messages
-					createMessage("user", "Can you provide a detailed example?", 100),
-					createMessage("assistant", "Here's a detailed example...", 200),
-				]
-
-				// Previous cache point placements
-				const previousCachePointPlacements: CachePointPlacement[] = [
-					{
-						index: 2,
-						type: "message",
-						tokensCovered: 400,
-					},
-					{
-						index: 4,
-						type: "message",
-						tokensCovered: 150,
-					},
-					{
-						index: 6,
-						type: "message",
-						tokensCovered: 150,
-					},
-				]
-
-				const config = createConfig({
-					modelInfo: multiPointModelInfo,
-					systemPrompt: "You are a helpful assistant.",
-					messages,
-					usePromptCache: true,
-					previousCachePointPlacements,
-				})
-
-				const strategy = new MultiPointStrategy(config)
-				const result = strategy.determineOptimalCachePoints()
-
-				// Verify we have cache points
-				expect(result.messageCachePointPlacements).toBeDefined()
-				expect(result.messageCachePointPlacements?.length).toBeGreaterThan(0)
-			})
-		})
-	})
-})

src/api/transform/cache-strategy/base-strategy.ts

@@ -1,172 +0,0 @@
-import { Anthropic } from "@anthropic-ai/sdk"
-import { ContentBlock, SystemContentBlock, Message, ConversationRole } from "@aws-sdk/client-bedrock-runtime"
-import { CacheStrategyConfig, CacheResult, CachePointPlacement } from "./types"
-
-export abstract class CacheStrategy {
-	/**
-	 * Determine optimal cache point placements and return the formatted result
-	 */
-	public abstract determineOptimalCachePoints(): CacheResult
-
-	protected config: CacheStrategyConfig
-	protected systemTokenCount: number = 0
-
-	constructor(config: CacheStrategyConfig) {
-		this.config = config
-		this.initializeMessageGroups()
-		this.calculateSystemTokens()
-	}
-
-	/**
-	 * Initialize message groups from the input messages
-	 */
-	protected initializeMessageGroups(): void {
-		if (!this.config.messages.length) return
-	}
-
-	/**
-	 * Calculate token count for system prompt using a more accurate approach
-	 */
-	protected calculateSystemTokens(): void {
-		if (this.config.systemPrompt) {
-			const text = this.config.systemPrompt
-
-			// Use a more accurate token estimation than simple character count
-			// Count words and add overhead for punctuation and special tokens
-			const words = text.split(/\s+/).filter((word) => word.length > 0)
-			// Average English word is ~1.3 tokens
-			let tokenCount = words.length * 1.3
-			// Add overhead for punctuation and special characters
-			tokenCount += (text.match(/[.,!?;:()[\]{}""''`]/g) || []).length * 0.3
-			// Add overhead for newlines
-			tokenCount += (text.match(/\n/g) || []).length * 0.5
-			// Add a small overhead for system prompt structure
-			tokenCount += 5
-
-			this.systemTokenCount = Math.ceil(tokenCount)
-		}
-	}
-
-	/**
-	 * Create a cache point content block
-	 */
-	protected createCachePoint(): ContentBlock {
-		return { cachePoint: { type: "default" } } as unknown as ContentBlock
-	}
-
-	/**
-	 * Convert messages to content blocks
-	 */
-	protected messagesToContentBlocks(messages: Anthropic.Messages.MessageParam[]): Message[] {
-		return messages.map((message) => {
-			const role: ConversationRole = message.role === "assistant" ? "assistant" : "user"
-
-			const content: ContentBlock[] = Array.isArray(message.content)
-				? message.content.map((block) => {
-						if (typeof block === "string") {
-							return { text: block } as unknown as ContentBlock
-						}
-						if ("text" in block) {
-							return { text: block.text } as unknown as ContentBlock
-						}
-						// Handle other content types if needed
-						return { text: "[Unsupported Content]" } as unknown as ContentBlock
-					})
-				: [{ text: message.content } as unknown as ContentBlock]
-
-			return {
-				role,
-				content,
-			}
-		})
-	}
-
-	/**
-	 * Check if a token count meets the minimum threshold for caching
-	 */
-	protected meetsMinTokenThreshold(tokenCount: number): boolean {
-		const minTokens = this.config.modelInfo.minTokensPerCachePoint
-		if (!minTokens) {
-			return false
-		}
-		return tokenCount >= minTokens
-	}
-
-	/**
-	 * Estimate token count for a message using a more accurate approach
-	 * This implementation is based on the BaseProvider's countTokens method
-	 * but adapted to work without requiring an instance of BaseProvider
-	 */
-	protected estimateTokenCount(message: Anthropic.Messages.MessageParam): number {
-		// Use a more sophisticated token counting approach
-		if (!message.content) return 0
-
-		let totalTokens = 0
-
-		if (Array.isArray(message.content)) {
-			for (const block of message.content) {
-				if (block.type === "text") {
-					// Use a more accurate token estimation than simple character count
-					// This is still an approximation but better than character/4
-					const text = block.text || ""
-					if (text.length > 0) {
-						// Count words and add overhead for punctuation and special tokens
-						const words = text.split(/\s+/).filter((word) => word.length > 0)
-						// Average English word is ~1.3 tokens
-						totalTokens += words.length * 1.3
-						// Add overhead for punctuation and special characters
-						totalTokens += (text.match(/[.,!?;:()[\]{}""''`]/g) || []).length * 0.3
-						// Add overhead for newlines
-						totalTokens += (text.match(/\n/g) || []).length * 0.5
-					}
-				} else if (block.type === "image") {
-					// For images, use a conservative estimate
-					totalTokens += 300
-				}
-			}
-		} else if (typeof message.content === "string") {
-			const text = message.content
-			// Count words and add overhead for punctuation and special tokens
-			const words = text.split(/\s+/).filter((word) => word.length > 0)
-			// Average English word is ~1.3 tokens
-			totalTokens += words.length * 1.3
-			// Add overhead for punctuation and special characters
-			totalTokens += (text.match(/[.,!?;:()[\]{}""''`]/g) || []).length * 0.3
-			// Add overhead for newlines
-			totalTokens += (text.match(/\n/g) || []).length * 0.5
-		}
-
-		// Add a small overhead for message structure
-		totalTokens += 10
-
-		return Math.ceil(totalTokens)
-	}
-
-	/**
-	 * Apply cache points to content blocks based on placements
-	 */
-	protected applyCachePoints(messages: Message[], placements: CachePointPlacement[]): Message[] {
-		const result: Message[] = []
-		for (let i = 0; i < messages.length; i++) {
-			const placement = placements.find((p) => p.index === i)
-
-			if (placement) {
-				messages[i].content?.push(this.createCachePoint())
-			}
-			result.push(messages[i])
-		}
-
-		return result
-	}
-
-	/**
-	 * Format the final result with cache points applied
-	 */
-	protected formatResult(systemBlocks: SystemContentBlock[] = [], messages: Message[]): CacheResult {
-		const result = {
-			system: systemBlocks,
-			messages,
-		}
-		return result
-	}
-}

src/api/transform/cache-strategy/multi-point-strategy.ts

@@ -1,314 +0,0 @@
-import { SystemContentBlock } from "@aws-sdk/client-bedrock-runtime"
-import { CacheStrategy } from "./base-strategy"
-import { CacheResult, CachePointPlacement } from "./types"
-import { logger } from "../../../utils/logging"
-
-/**
- * Strategy for handling multiple cache points.
- * Creates cache points after messages as soon as uncached tokens exceed minimumTokenCount.
- */
-export class MultiPointStrategy extends CacheStrategy {
-	/**
-	 * Determine optimal cache point placements and return the formatted result
-	 */
-	public determineOptimalCachePoints(): CacheResult {
-		// If prompt caching is disabled or no messages, return without cache points
-		if (!this.config.usePromptCache || this.config.messages.length === 0) {
-			return this.formatWithoutCachePoints()
-		}
-
-		const supportsSystemCache = this.config.modelInfo.cachableFields.includes("system")
-		const supportsMessageCache = this.config.modelInfo.cachableFields.includes("messages")
-		const minTokensPerPoint = this.config.modelInfo.minTokensPerCachePoint
-		let remainingCachePoints: number = this.config.modelInfo.maxCachePoints
-
-		// First, determine if we'll use a system cache point
-		const useSystemCache =
-			supportsSystemCache && this.config.systemPrompt && this.meetsMinTokenThreshold(this.systemTokenCount)
-
-		// Handle system blocks
-		let systemBlocks: SystemContentBlock[] = []
-		if (this.config.systemPrompt) {
-			systemBlocks = [{ text: this.config.systemPrompt } as unknown as SystemContentBlock]
-			if (useSystemCache) {
-				systemBlocks.push(this.createCachePoint() as unknown as SystemContentBlock)
-				remainingCachePoints--
-			}
-		}
-
-		// If message caching isn't supported, return with just system caching
-		if (!supportsMessageCache) {
-			return this.formatResult(systemBlocks, this.messagesToContentBlocks(this.config.messages))
-		}
-
-		const placements = this.determineMessageCachePoints(minTokensPerPoint, remainingCachePoints)
-		const messages = this.messagesToContentBlocks(this.config.messages)
-		let cacheResult = this.formatResult(systemBlocks, this.applyCachePoints(messages, placements))
-
-		// Store the placements for future use (to maintain consistency across consecutive messages)
-		// This needs to be handled by the caller by passing these placements back in the next call
-		cacheResult.messageCachePointPlacements = placements
-
-		return cacheResult
-	}
-
-	/**
-	 * Determine optimal cache point placements for messages
-	 * This method handles both new conversations and growing conversations
-	 *
-	 * @param minTokensPerPoint Minimum tokens required per cache point
-	 * @param remainingCachePoints Number of cache points available
-	 * @returns Array of cache point placements
-	 */
-	private determineMessageCachePoints(
-		minTokensPerPoint: number,
-		remainingCachePoints: number,
-	): CachePointPlacement[] {
-		if (this.config.messages.length <= 1) {
-			return []
-		}
-
-		const placements: CachePointPlacement[] = []
-		const totalMessages = this.config.messages.length
-		const previousPlacements = this.config.previousCachePointPlacements || []
-
-		// Special case: If previousPlacements is empty, place initial cache points
-		if (previousPlacements.length === 0) {
-			let currentIndex = 0
-
-			while (currentIndex < totalMessages && remainingCachePoints > 0) {
-				const newPlacement = this.findOptimalPlacementForRange(
-					currentIndex,
-					totalMessages - 1,
-					minTokensPerPoint,
-				)
-
-				if (newPlacement) {
-					placements.push(newPlacement)
-					currentIndex = newPlacement.index + 1
-					remainingCachePoints--
-				} else {
-					break
-				}
-			}
-
-			return placements
-		}
-
-		// Calculate tokens in new messages (added since last cache point placement)
-		const lastPreviousIndex = previousPlacements[previousPlacements.length - 1].index
-		const newMessagesTokens = this.config.messages
-			.slice(lastPreviousIndex + 1)
-			.reduce((acc, curr) => acc + this.estimateTokenCount(curr), 0)
-
-		// If new messages have enough tokens for a cache point, we need to decide
-		// whether to keep all previous cache points or combine some
-		if (newMessagesTokens >= minTokensPerPoint) {
-			// If we have enough cache points for all previous placements plus a new one, keep them all
-			if (remainingCachePoints > previousPlacements.length) {
-				// Keep all previous placements
-				for (const placement of previousPlacements) {
-					if (placement.index < totalMessages) {
-						placements.push(placement)
-					}
-				}
-
-				// Add a new placement for the new messages
-				const newPlacement = this.findOptimalPlacementForRange(
-					lastPreviousIndex + 1,
-					totalMessages - 1,
-					minTokensPerPoint,
-				)
-
-				if (newPlacement) {
-					placements.push(newPlacement)
-				}
-			} else {
-				// We need to decide which previous cache points to keep and which to combine
-				// Strategy: Compare the token count of new messages with the smallest combined token gap
-
-				// First, analyze the token distribution between previous cache points
-				const tokensBetweenPlacements: number[] = []
-				let startIdx = 0
-
-				for (const placement of previousPlacements) {
-					const tokens = this.config.messages
-						.slice(startIdx, placement.index + 1)
-						.reduce((acc, curr) => acc + this.estimateTokenCount(curr), 0)
-
-					tokensBetweenPlacements.push(tokens)
-					startIdx = placement.index + 1
-				}
-
-				// Find the two consecutive placements with the smallest token gap
-				let smallestGapIndex = 0
-				let smallestGap = Number.MAX_VALUE
-
-				for (let i = 0; i < tokensBetweenPlacements.length - 1; i++) {
-					const gap = tokensBetweenPlacements[i] + tokensBetweenPlacements[i + 1]
-					if (gap < smallestGap) {
-						smallestGap = gap
-						smallestGapIndex = i
-					}
-				}
-
-				// Only combine cache points if it's beneficial
-				// Compare the token count of new messages with the smallest combined token gap
-				// Apply a required percentage increase to ensure reallocation is worth it
-				const requiredPercentageIncrease = 1.2 // 20% increase required
-				const requiredTokenThreshold = smallestGap * requiredPercentageIncrease
-
-				if (newMessagesTokens >= requiredTokenThreshold) {
-					// It's beneficial to combine cache points since new messages have significantly more tokens
-					logger.info("Combining cache points is beneficial", {
-						ctx: "cache-strategy",
-						newMessagesTokens,
-						smallestGap,
-						requiredTokenThreshold,
-						action: "combining_cache_points",
-					})
-
-					// Combine the two placements with the smallest gap
-					for (let i = 0; i < previousPlacements.length; i++) {
-						if (i !== smallestGapIndex && i !== smallestGapIndex + 1) {
-							// Keep this placement
-							if (previousPlacements[i].index < totalMessages) {
-								placements.push(previousPlacements[i])
-							}
-						} else if (i === smallestGapIndex) {
-							// Replace with a combined placement
-							const combinedEndIndex = previousPlacements[i + 1].index
-
-							// Find the optimal placement within this combined range
-							const startOfRange = i === 0 ? 0 : previousPlacements[i - 1].index + 1
-							const combinedPlacement = this.findOptimalPlacementForRange(
-								startOfRange,
-								combinedEndIndex,
-								minTokensPerPoint,
-							)
-
-							if (combinedPlacement) {
-								placements.push(combinedPlacement)
-							}
-
-							// Skip the next placement as we've combined it
-							i++
-						}
-					}
-
-					// If we freed up a cache point, use it for the new messages
-					if (placements.length < remainingCachePoints) {
-						const newPlacement = this.findOptimalPlacementForRange(
-							lastPreviousIndex + 1,
-							totalMessages - 1,
-							minTokensPerPoint,
-						)
-
-						if (newPlacement) {
-							placements.push(newPlacement)
-						}
-					}
-				} else {
-					// It's not beneficial to combine cache points
-					// Keep all previous placements and don't add a new one for the new messages
-					logger.info("Combining cache points is not beneficial", {
-						ctx: "cache-strategy",
-						newMessagesTokens,
-						smallestGap,
-						action: "keeping_existing_cache_points",
-					})
-
-					// Keep all previous placements that are still valid
-					for (const placement of previousPlacements) {
-						if (placement.index < totalMessages) {
-							placements.push(placement)
-						}
-					}
-				}
-			}
-
-			return placements
-		} else {
-			// New messages don't have enough tokens for a cache point
-			// Keep all previous placements that are still valid
-			for (const placement of previousPlacements) {
-				if (placement.index < totalMessages) {
-					placements.push(placement)
-				}
-			}
-
-			return placements
-		}
-	}
-
-	/**
-	 * Find the optimal placement for a cache point within a specified range of messages
-	 * Simply finds the last user message in the range
-	 */
-	private findOptimalPlacementForRange(
-		startIndex: number,
-		endIndex: number,
-		minTokensPerPoint: number,
-	): CachePointPlacement | null {
-		if (startIndex >= endIndex) {
-			return null
-		}
-
-		// Find the last user message in the range
-		let lastUserMessageIndex = -1
-		for (let i = endIndex; i >= startIndex; i--) {
-			if (this.config.messages[i].role === "user") {
-				lastUserMessageIndex = i
-				break
-			}
-		}
-
-		if (lastUserMessageIndex >= 0) {
-			// Calculate the total tokens covered from the previous cache point (or start of conversation)
-			// to this cache point. This ensures tokensCovered represents the full span of tokens
-			// that will be cached by this cache point.
-			let totalTokensCovered = 0
-
-			// Find the previous cache point index
-			const previousPlacements = this.config.previousCachePointPlacements || []
-			let previousCachePointIndex = -1
-
-			for (const placement of previousPlacements) {
-				if (placement.index < startIndex && placement.index > previousCachePointIndex) {
-					previousCachePointIndex = placement.index
-				}
-			}
-
-			// Calculate tokens from previous cache point (or start) to this cache point
-			const tokenStartIndex = previousCachePointIndex + 1
-			totalTokensCovered = this.config.messages
-				.slice(tokenStartIndex, lastUserMessageIndex + 1)
-				.reduce((acc, curr) => acc + this.estimateTokenCount(curr), 0)
-
-			// Guard clause: ensure we have enough tokens to justify a cache point
-			if (totalTokensCovered < minTokensPerPoint) {
-				return null
-			}
-			return {
-				index: lastUserMessageIndex,
-				type: "message",
-				tokensCovered: totalTokensCovered,
-			}
-		}
-
-		return null
-	}
-
-	/**
-	 * Format result without cache points
-	 *
-	 * @returns Cache result without cache points
-	 */
-	private formatWithoutCachePoints(): CacheResult {
-		const systemBlocks: SystemContentBlock[] = this.config.systemPrompt
-			? [{ text: this.config.systemPrompt } as unknown as SystemContentBlock]
-			: []
-
-		return this.formatResult(systemBlocks, this.messagesToContentBlocks(this.config.messages))
-	}
-}

src/api/transform/cache-strategy/types.ts

@@ -1,68 +0,0 @@
-import { Anthropic } from "@anthropic-ai/sdk"
-import { SystemContentBlock, Message } from "@aws-sdk/client-bedrock-runtime"
-
-/**
- * Information about a model's capabilities and constraints
- */
-export interface ModelInfo {
-	/** Maximum number of tokens the model can generate */
-	maxTokens: number
-	/** Maximum context window size in tokens */
-	contextWindow: number
-	/** Whether the model supports prompt caching */
-	supportsPromptCache: boolean
-	/** Maximum number of cache points supported by the model */
-	maxCachePoints: number
-	/** Minimum number of tokens required for a cache point */
-	minTokensPerCachePoint: number
-	/** Fields that can be cached */
-	cachableFields: Array<"system" | "messages" | "tools">
-}
-
-/**
- * Cache point definition
- */
-export interface CachePoint {
-	/** Type of cache point */
-	type: "default"
-}
-
-/**
- * Result of cache strategy application
- */
-export interface CacheResult {
-	/** System content blocks */
-	system: SystemContentBlock[]
-	/** Message content blocks */
-	messages: Message[]
-	/** Cache point placements for messages (for maintaining consistency across consecutive messages) */
-	messageCachePointPlacements?: CachePointPlacement[]
-}
-
-/**
- * Represents the position and metadata for a cache point
- */
-export interface CachePointPlacement {
-	/** Where to insert the cache point */
-	index: number
-	/** Type of cache point */
-	type: "system" | "message"
-	/** Number of tokens this cache point covers */
-	tokensCovered: number
-}
-
-/**
- * Configuration for the caching strategy
- */
-export interface CacheStrategyConfig {
-	/** Model information */
-	modelInfo: ModelInfo
-	/** System prompt text */
-	systemPrompt?: string
-	/** Messages to process */
-	messages: Anthropic.Messages.MessageParam[]
-	/** Whether to use prompt caching */
-	usePromptCache: boolean
-	/** Previous cache point placements (for maintaining consistency across consecutive messages) */
-	previousCachePointPlacements?: CachePointPlacement[]
-}

src/package.json

@@ -450,6 +450,7 @@
 		"clean": "rimraf README.md CHANGELOG.md LICENSE dist logs mock .turbo"
 	},
 	"dependencies": {
+		"@ai-sdk/amazon-bedrock": "^4.0.50",
 		"@ai-sdk/cerebras": "^1.0.0",
 		"@ai-sdk/deepseek": "^2.0.14",
 		"@ai-sdk/fireworks": "^2.0.26",
@@ -458,8 +459,6 @@
 		"@ai-sdk/groq": "^3.0.19",
 		"@ai-sdk/mistral": "^3.0.0",
 		"@ai-sdk/xai": "^3.0.46",
-		"sambanova-ai-provider": "^1.2.2",
-		"@anthropic-ai/bedrock-sdk": "^0.10.2",
 		"@anthropic-ai/sdk": "^0.37.0",
 		"@anthropic-ai/vertex-sdk": "^0.7.0",
 		"@aws-sdk/client-bedrock-runtime": "^3.922.0",
@@ -518,6 +517,7 @@
 		"puppeteer-core": "^23.4.0",
 		"reconnecting-eventsource": "^1.6.4",
 		"safe-stable-stringify": "^2.5.0",
+		"sambanova-ai-provider": "^1.2.2",
 		"sanitize-filename": "^1.6.3",
 		"say": "^0.16.0",
 		"semver-compare": "^1.0.0",