aiproxy/config.example.yaml

# AIProxy Configuration File
# Priority: Environment Variables > Config File > Database
# This file allows you to configure channels, model configs, and options without using the database

# Channels Configuration
# Note: Channels from YAML are assigned negative IDs automatically and are not persisted to database
# They are merged with database channels in memory
channels:
  - name: "openai-channel-1"
    type_name: "openai"  # Can use type_name instead of numeric type
    key: "sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
    base_url: "https://api.openai.com"
    models:
      - "gpt-4"
      - "gpt-3.5-turbo"
    model_mapping:
      "gpt-4": "gpt-4-0613"
    status: 1  # 1=Enabled, 2=Disabled
    priority: 0
    balance: 100.0
    balance_threshold: 10.0
    enabled_auto_balance_check: true
    sets:
      - "default"
    config:
      spec:
        organization: "org-xxxxx"

  - name: "azure-channel-1"
    type_name: "azure"  # Type name is case-insensitive
    key: "your-azure-api-key"
    base_url: "https://your-resource.openai.azure.com"
    models:
      - "gpt-4"
    status: 1
    priority: 1
    sets:
      - "default"

  - name: "claude-channel-1"
    type_name: "claude"  # "claude" is an alias for "anthropic"
    key: "sk-ant-xxxxx"
    base_url: "https://api.anthropic.com"
    models:
      - "claude-3-opus-20240229"
      - "claude-3-sonnet-20240229"
    status: 1
    priority: 0
    sets:
      - "default"
      - "premium"

  # You can also use numeric type if preferred
  - name: "gemini-channel-1"
    type: 24  # Google Gemini
    key: "your-gemini-api-key"
    models:
      - "gemini-pro"
    status: 1

# Model Configurations
modelconfigs:
  - model: "gpt-4"
    owner: "openai"
    type_name: "chat"  # Can use type_name instead of numeric type
    # OR use numeric type:
    # type: 1  # ChatCompletions
    rpm: 3500
    tpm: 80000
    retry_times: 3
    timeout_config:
      request_timeout: 300
      stream_request_timeout: 600
    warn_error_rate: 0.5
    max_error_rate: 0.8
    price:
      input: 0.03
      output: 0.06
    config:
      max_context_tokens: 8192
      max_output_tokens: 4096
      vision: false
      tool_choice: true

  - model: "gpt-3.5-turbo"
    owner: "openai"
    type_name: "chat"
    rpm: 3500
    tpm: 90000
    price:
      input: 0.0005
      output: 0.0015
    config:
      max_context_tokens: 16384
      max_output_tokens: 4096

  - model: "claude-3-opus-20240229"
    owner: "anthropic"
    type_name: "chat"
    rpm: 4000
    tpm: 400000
    price:
      input: 0.015
      output: 0.075
    config:
      max_context_tokens: 200000
      max_output_tokens: 4096
      vision: true

  - model: "claude-3-sonnet-20240229"
    owner: "anthropic"
    type_name: "chat"
    rpm: 4000
    tpm: 400000
    price:
      input: 0.003
      output: 0.015
    config:
      max_context_tokens: 200000
      max_output_tokens: 4096
      vision: true

  - model: "text-embedding-3-small"
    owner: "openai"
    type_name: "embedding"  # Embedding model
    rpm: 3000
    tpm: 1000000
    price:
      input: 0.00002
      output: 0
    config:
      max_context_tokens: 8191

  - model: "dall-e-3"
    owner: "openai"
    type_name: "image"  # Image generation
    rpm: 50
    image_quality_prices:
      "1024x1024":
        "standard": 0.040
        "hd": 0.080
      "1024x1792":
        "standard": 0.080
        "hd": 0.120
      "1792x1024":
        "standard": 0.080
        "hd": 0.120

# System Options Configuration
options:
  # Log retention settings (in hours)
  LogStorageHours: "168"  # 7 days
  RetryLogStorageHours: "72"  # 3 days
  LogDetailStorageHours: "24"  # 1 day

  # Clean log batch size
  CleanLogBatchSize: "1000"

  # IP rate limiting
  IPGroupsThreshold: "100"  # Requests per minute
  IPGroupsBanThreshold: "200"  # Ban threshold

  # Log detail settings
  SaveAllLogDetail: "false"
  LogDetailRequestBodyMaxSize: "10000"
  LogDetailResponseBodyMaxSize: "10000"

  # Retry settings
  RetryTimes: "3"

  # Group settings
  GroupMaxTokenNum: "0"  # 0 means unlimited
  GroupConsumeLevelRatio: '{"1":1,"2":0.9,"3":0.8}'

  # Error rate alerts
  DefaultWarnNotifyErrorRate: "0.5"

  # Usage alerts
  UsageAlertThreshold: "100"
  UsageAlertMinAvgThreshold: "10"

  # Fuzzy token threshold
  FuzzyTokenThreshold: "240000"

  # Disable serve (for maintenance)
  DisableServe: "false"