models:
  - model_id: mock.default
    provider_name: mock
    model_name: mock-echo
    temperature: 0
    max_tokens: 384
    timeout: 10
    retry:
      max_attempts: 1
      initial_delay_seconds: 0.5
      max_delay_seconds: 1
      backoff_multiplier: 1
      jitter_ratio: 0
    rate_limits:
      max_concurrency: 100
    provider_kwargs:
      prefix: "[MOCK]"
  - model_id: mock.judge
    provider_name: mock
    model_name: mock-judge
    temperature: 0
    max_tokens: 512
    timeout: 10
    retry:
      max_attempts: 1
      initial_delay_seconds: 0.5
      max_delay_seconds: 1
      backoff_multiplier: 1
      jitter_ratio: 0
    rate_limits:
      max_concurrency: 100
    provider_kwargs:
      mode: judge_json
  - model_id: qwen-test
    provider_name: openai_compatible
    model_name: qwen
    api_base: https://o9peaj85dhhcchhhheakbppg8mj8cboj.openapi-qb-ai.sii.edu.cn/v1
    api_key_env: QWEN_TEST_API_KEY
    temperature: 0
    max_tokens: 1024
    timeout: 120
    retry:
      max_attempts: 3
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 64
    supports_system_prompt: true
    supports_stream: true
    supports_reasoning: true
    provider_kwargs:
      chat_template_kwargs:
        enable_thinking: false
      runtime_max_tokens: 512
      served_model_name: qwen3.6-35B-sft-v1
  - model_id: qwen3.6-35B-sft-v1
    provider_name: openai_compatible
    model_name: qwen
    api_base: https://o9peaj85dhhcchhhheakbppg8mj8cboj.openapi-qb-ai.sii.edu.cn/v1
    api_key_env: QWEN_TEST_API_KEY
    temperature: 0
    max_tokens: 1024
    timeout: 120
    retry:
      max_attempts: 3
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
      retryable_status_codes: [408, 409, 429, 500, 502, 503, 504]
    rate_limits:
      max_concurrency: 4
    supports_system_prompt: true
    supports_stream: true
    supports_reasoning: true
    provider_kwargs:
      chat_template_kwargs:
        enable_thinking: false
      served_model_name: qwen3.6-35B-sft-v1
  - model_id: gpt-5.4
    provider_name: openai_compatible
    model_name: gpt-5.4
    api_base: http://35.220.164.252:3888/v1
    api_key_env: EXTERNAL_SUPPORT_API_KEY
    temperature: 0
    max_tokens: 4096
    timeout: 120
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: gpt-5.2-pro
    provider_name: openai_compatible
    model_name: gpt-5.2-pro
    api_base: http://35.220.164.252:3888/v1
    api_key_env: EXTERNAL_SUPPORT_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: claude-sonnet-4-6
    provider_name: openai_compatible
    model_name: claude-sonnet-4-6
    api_base: https://gpt-agent.cc/v1
    api_key_env: GPT_AGENT_API_KEY
    temperature: 0
    max_tokens: 4096
    timeout: 120
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: claude-opus-4-6
    provider_name: openai_compatible
    model_name: claude-opus-4-6
    api_base: https://gpt-agent.cc/v1
    api_key_env: GPT_AGENT_API_KEY
    temperature: 0
    max_tokens: 4096
    timeout: 120
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: gemini-3.5-flash
    provider_name: openai_compatible
    model_name: gemini-3.5-flash
    api_base: http://35.220.164.252:3888/v1
    api_key_env: EXTERNAL_SUPPORT_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: deepseek-v4-pro
    provider_name: openai_compatible
    model_name: deepseek-v4-pro
    api_base: https://api.deepseek.com
    api_key_env: DEEPSEEK_OFFICIAL_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: deepseek-v4-flash
    provider_name: openai_compatible
    model_name: deepseek-v4-flash
    api_base: https://api.deepseek.com
    api_key_env: DEEPSEEK_OFFICIAL_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: glm-5.1
    provider_name: openai_compatible
    model_name: glm-5.1
    api_base: https://open.bigmodel.cn/api/paas/v4
    api_key_env: ZHIPU_OFFICIAL_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 10
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: claude-opus-4-8
    provider_name: openai_compatible
    model_name: claude-opus-4-8
    api_base: https://gpt-agent.cc/v1
    api_key_env: GPT_AGENT_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 360
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: doubao-seed-2.0-pro
    provider_name: openai_compatible
    model_name: doubao-seed-2.0-pro
    api_base: https://gpt-agent.cc/v1
    api_key_env: GPT_AGENT_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: deepseek-v3.2
    provider_name: openai_compatible
    model_name: deepseek-v3.2
    api_base: https://api.innospark.cn/v1
    api_key_env: INNOSPARK_RELAY_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: doubao-seed-2-0-pro-260215
    provider_name: openai_compatible
    model_name: doubao-seed-2-0-pro-260215
    api_base: https://api.innospark.cn/v1
    api_key_env: INNOSPARK_RELAY_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: kimi-k2.6
    provider_name: openai_compatible
    model_name: kimi-k2.6
    api_base: https://api.innospark.cn/v1
    api_key_env: INNOSPARK_RELAY_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 3
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      rpm: 8
      max_concurrency: 1
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: gemini-3-flash-preview
    provider_name: openai_compatible
    model_name: gemini-3-flash-preview
    api_base: https://api.innospark.cn/v1
    api_key_env: INNOSPARK_RELAY_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 3
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: deepseek-r1-250528
    provider_name: openai_compatible
    model_name: deepseek-r1-250528
    api_base: https://api.innospark.cn/v1
    api_key_env: INNOSPARK_RELAY_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 480
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
      retryable_status_codes: [408, 409, 429, 500, 502, 503, 504]
    rate_limits:
      max_concurrency: 2
    supports_system_prompt: true
    supports_stream: true
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: gemini-3.1-pro-preview
    provider_name: openai_compatible
    model_name: gemini-3.1-pro-preview
    api_base: https://api.innospark.cn/v1
    api_key_env: INNOSPARK_RELAY_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 2
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
    rate_limits:
      max_concurrency: 100
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: true
    provider_kwargs: {}
  - model_id: innospark-235b
    provider_name: openai_compatible
    model_name: InnoSpark-235B
    api_base: https://innospark-api.aiecnu.net/v1
    api_key_env: INNOSPARK_AIECNU_API_KEY
    temperature: 0
    max_tokens: 1024
    timeout: 600
    retry:
      max_attempts: 2
      initial_delay_seconds: 5
      max_delay_seconds: 120
      backoff_multiplier: 2
      jitter_ratio: 0.2
      retryable_status_codes: [408, 409, 429, 500, 502, 503, 504]
    rate_limits:
      max_concurrency: 1
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: false
    provider_kwargs:
      runtime_max_tokens: 512
  - model_id: qwen3.5-397b-a17b
    provider_name: openai_compatible
    model_name: qwen3.5-397b-a17b
    api_base: https://dashscope.aliyuncs.com/compatible-mode/v1
    api_key_env: DASHSCOPE_API_KEY
    temperature: 0
    max_tokens: 8192
    timeout: 240
    retry:
      max_attempts: 3
      initial_delay_seconds: 1
      max_delay_seconds: 20
      backoff_multiplier: 2
      jitter_ratio: 0.2
      retryable_status_codes: [408, 409, 429, 500, 502, 503, 504]
    rate_limits:
      max_concurrency: 10
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: false
    provider_kwargs:
      # DashScope only honors top-level enable_thinking for this model;
      # chat_template_kwargs.enable_thinking is IGNORED here (verified empirically),
      # so we must send it at the top level. provider_kwargs is merged into the
      # request payload top level by the openai_compatible adapter.
      enable_thinking: false
  - model_id: safe-innospark
    provider_name: openai_compatible
    model_name: Safe-InnoSpark
    api_base: https://innospark-api.aiecnu.net/v1
    api_key_env: INNOSPARK_AIECNU_API_KEY
    temperature: 0
    max_tokens: 1024
    timeout: 600
    retry:
      max_attempts: 2
      initial_delay_seconds: 5
      max_delay_seconds: 120
      backoff_multiplier: 2
      jitter_ratio: 0.2
      retryable_status_codes: [408, 409, 429, 500, 502, 503, 504]
    rate_limits:
      max_concurrency: 1
    supports_system_prompt: true
    supports_stream: false
    supports_reasoning: false
    provider_kwargs:
      runtime_max_tokens: 512