Legacy Providers Configuration

Old providers config format (deprecated, use model_list instead)

The providers configuration format is deprecated. Use the new model_list format instead for better flexibility and zero-code provider addition.

Overview

The legacy providers configuration was PicoClaw’s original way to configure LLM providers. It’s still supported for backward compatibility, but new configurations should use model_list.

Legacy Configuration Format

{
  "providers": {
    "zhipu": {
      "api_key": "your-key",
      "api_base": "https://open.bigmodel.cn/api/paas/v4"
    },
    "openai": {
      "api_key": "sk-...",
      "api_base": "https://api.openai.com/v1"
    },
    "anthropic": {
      "api_key": "sk-ant-...",
      "api_base": "https://api.anthropic.com/v1"
    }
  },
  "agents": {
    "defaults": {
      "provider": "zhipu",
      "model": "glm-4.7"
    }
  }
}

Supported Legacy Providers

• anthropic - Anthropic Claude
• openai - OpenAI GPT
• litellm - LiteLLM Proxy
• openrouter - OpenRouter
• groq - Groq
• zhipu - Zhipu AI (智谱)
• vllm - vLLM
• gemini - Google Gemini
• nvidia - NVIDIA
• ollama - Ollama (local)
• moonshot - Moonshot AI
• shengsuanyun - ShengsuanYun
• deepseek - DeepSeek
• cerebras - Cerebras
• volcengine - Volcengine (火山引擎)
• github_copilot - GitHub Copilot
• antigravity - Google Cloud Code Assist
• qwen - Alibaba Qwen
• mistral - Mistral AI

Migration to model_list

Example 1: Single Provider

Old format:

{
  "providers": {
    "zhipu": {
      "api_key": "your-key",
      "api_base": "https://open.bigmodel.cn/api/paas/v4"
    }
  },
  "agents": {
    "defaults": {
      "provider": "zhipu",
      "model": "glm-4.7"
    }
  }
}

New format:

{
  "model_list": [
    {
      "model_name": "glm-4.7",
      "model": "zhipu/glm-4.7",
      "api_key": "your-key",
      "api_base": "https://open.bigmodel.cn/api/paas/v4"
    }
  ],
  "agents": {
    "defaults": {
      "model_name": "glm-4.7"
    }
  }
}

Example 2: Multiple Providers

Old format:

{
  "providers": {
    "openai": {
      "api_key": "sk-openai-key"
    },
    "anthropic": {
      "api_key": "sk-ant-key"
    }
  },
  "agents": {
    "defaults": {
      "provider": "anthropic",
      "model": "claude-sonnet-4.6"
    }
  }
}

New format:

{
  "model_list": [
    {
      "model_name": "gpt-5.2",
      "model": "openai/gpt-5.2",
      "api_key": "sk-openai-key"
    },
    {
      "model_name": "claude-sonnet-4.6",
      "model": "anthropic/claude-sonnet-4.6",
      "api_key": "sk-ant-key"
    }
  ],
  "agents": {
    "defaults": {
      "model_name": "claude-sonnet-4.6"
    }
  }
}

Automatic Migration

PicoClaw automatically migrates legacy configurations at runtime:

1. If model_list is empty and providers has configuration
2. PicoClaw converts providers to model_list internally
3. Your config file is not modified
4. You can manually update to the new format at any time

Environment Variables (Legacy)

Legacy provider environment variables still work:

# Zhipu
PICOCLAW_PROVIDERS_ZHIPU_API_KEY=your-key
PICOCLAW_PROVIDERS_ZHIPU_API_BASE=https://open.bigmodel.cn/api/paas/v4

# OpenAI
PICOCLAW_PROVIDERS_OPENAI_API_KEY=sk-...
PICOCLAW_PROVIDERS_OPENAI_API_BASE=https://api.openai.com/v1

# Anthropic
PICOCLAW_PROVIDERS_ANTHROPIC_API_KEY=sk-ant-...

Provider-Specific Options

Request Timeout

Set custom timeout for provider requests (seconds):

{
  "providers": {
    "openai": {
      "api_key": "sk-...",
      "request_timeout": 300
    }
  }
}

Proxy Configuration

Configure HTTP proxy for provider:

{
  "providers": {
    "openai": {
      "api_key": "sk-...",
      "proxy": "http://proxy.example.com:8080"
    }
  }
}

OpenAI Web Search

Enable web search augmentation for OpenAI:

{
  "providers": {
    "openai": {
      "api_key": "sk-...",
      "web_search": true
    }
  }
}

GitHub Copilot Connect Mode

{
  "providers": {
    "github_copilot": {
      "connect_mode": "grpc",
      "api_base": "http://localhost:4321"
    }
  }
}

Values: stdio or grpc

Why Migrate?

The new model_list format provides:

1. Multiple models per provider - Configure GPT-5.2 and GPT-4o separately
2. Load balancing - Multiple endpoints for the same model
3. Better fallbacks - Explicit model fallback chains
4. Zero-code providers - Add OpenAI-compatible APIs without code changes
5. Cleaner configuration - Model-centric instead of provider-centric

Next Steps

• Model Configuration (New Format)
• Agent Configuration
• Environment Variables