{ "user_patches": [ { "name": "custom_temperature_limit", "description": "Limit temperature to 1.5 for all models", "conditions": [ { "key": "temperature", "operator": "exists", "value": "" } ], "operations": [ { "op": "limit", "key": "temperature", "value": 1.5 } ] }, { "name": "add_default_top_p", "description": "Add default top_p if not specified", "conditions": [ { "key": "top_p", "operator": "not_exists", "value": "" } ], "operations": [ { "op": "add", "key": "top_p", "value": 0.9 } ] }, { "name": "remove_system_fingerprint_for_old_models", "description": "Remove system_fingerprint for models that don't support it", "conditions": [ { "key": "model", "operator": "regex", "value": "gpt-3\\.5-turbo" }, { "key": "system_fingerprint", "operator": "exists", "value": "" } ], "operations": [ { "op": "delete", "key": "system_fingerprint" } ] }, { "name": "anthropic_specific_settings", "description": "Apply Anthropic-specific optimizations", "conditions": [ { "key": "model", "operator": "contains", "value": "claude" } ], "operations": [ { "op": "limit", "key": "max_tokens", "value": 4096 }, { "op": "set", "key": "anthropic_version", "value": "2023-06-01" } ] }, { "name": "convert_legacy_stop_to_stop_sequences", "description": "Convert stop parameter to stop_sequences for Claude models", "conditions": [ { "key": "model", "operator": "contains", "value": "claude" }, { "key": "stop", "operator": "exists", "value": "" } ], "operations": [ { "op": "set", "key": "stop_sequences", "value": "{{stop}}" }, { "op": "delete", "key": "stop" } ] }, { "name": "nested_parameter_example", "description": "Example of setting nested parameters", "conditions": [ { "key": "model", "operator": "equals", "value": "example-model" } ], "operations": [ { "op": "set", "key": "generation_config.max_output_tokens", "value": 2048 }, { "op": "set", "key": "generation_config.temperature", "value": 0.7 }, { "op": "set", "key": "safety_settings.category", "value": "HARM_CATEGORY_DANGEROUS_CONTENT" } ] }, { "name": "model_specific_token_limits", "description": "Set appropriate token limits based on model capabilities", "conditions": [ { "key": "original_model", "operator": "regex", "value": "gpt-4-turbo" } ], "operations": [ { "op": "limit", "key": "max_tokens", "value": 4096 } ] } ] }