Configuration

Learn how to configure Weaver for AI agent orchestration with providers, tools, and channels

Weaver uses a flexible configuration system that combines JSON config files with environment variables. This guide covers all configuration options for agent defaults, LLM providers, chat channels, and tools.

Configuration Files

Weaver stores configuration in ~/.weaver/config.json. You can also use environment variables to override any JSON setting.

Default Configuration

When you run weaver onboard, a default configuration is created:

{
  "agents": {
    "defaults": {
      "workspace": "/root/.weaver/workspace",
      "restrict_to_workspace": true,
      "model": "gemini-3-flash-preview",
      "max_tokens": 8192,
      "temperature": 0.7,
      "max_tool_iterations": 20
    }
  },
  "providers": {
    "gemini": {
      "api_key": "YOUR_GEMINI_API_KEY"
    }
  },
  "tools": {
    "web": {
      "search": {
        "api_key": "",
        "max_results": 5
      }
    }
  },
  "heartbeat": {
    "enabled": true,
    "interval": 30
  },
  "gateway": {
    "host": "0.0.0.0",
    "port": 18790
  }
}

Agent Configuration

Configure default settings for all spawned agents:

1. Set Workspace Directory

   Define where agents can read and write files:

   {
     "agents": {
       "defaults": {
         "workspace": "~/.weaver/workspace",
         "restrict_to_workspace": true
       }
     }
   }
   

   Set restrict_to_workspace to false if you want agents to access files outside the workspace (use with caution).

2. Configure Model Settings

   Set the default LLM model and parameters:

   {
     "agents": {
       "defaults": {
         "provider": "gemini",
         "model": "gemini-3-flash-preview",
         "max_tokens": 8192,
         "temperature": 0.7,
         "max_tool_iterations": 20
       }
     }
   }
   

   Parameters:

   • provider: The LLM provider to use (gemini, openai, anthropic, etc.)
   • model: Model identifier for the provider
   • max_tokens: Maximum tokens in the response
   • temperature: Randomness in responses (0.0-1.0)
   • max_tool_iterations: Maximum tool execution loops before stopping
3. Override with Environment Variables

   You can override any agent default with environment variables:

   export WEAVER_AGENTS_DEFAULTS_MODEL="claude-3-5-sonnet-20241022"
   export WEAVER_AGENTS_DEFAULTS_TEMPERATURE="0.5"
   export WEAVER_AGENTS_DEFAULTS_MAX_TOKENS="16384"
   

LLM Providers

Weaver supports multiple LLM providers. Configure the ones you need:

Supported Providers

Anthropic

{
  "providers": {
    "anthropic": {
      "api_key": "sk-ant-xxx",
      "api_base": "",
      "proxy": ""
    }
  }
}

Using Environment Variables

For security, use environment variables for API keys:

.env

# LLM Provider
OPENROUTER_API_KEY=sk-or-v1-xxx
ZHIPU_API_KEY=xxx
ANTHROPIC_API_KEY=sk-ant-xxx
OPENAI_API_KEY=sk-xxx
GEMINI_API_KEY=xxx

Weaver automatically reads from .env files or system environment variables.

Tools Configuration

Configure tools that agents can use:

Web Search Tools

Brave Search

{
  "tools": {
    "web": {
      "brave": {
        "enabled": true,
        "api_key": "BSA...",
        "max_results": 5
      }
    }
  }
}

Environment variable:

export BRAVE_SEARCH_API_KEY=BSA...

Gateway Configuration

Configure the Weaver gateway server for chat channels:

{
  "gateway": {
    "host": "0.0.0.0",
    "port": 18790
  }
}

The gateway provides:

• HTTP health check endpoint at /health
• Chat channel webhooks
• Agent communication bus

Heartbeat Service

Keep long-running sessions alive:

{
  "heartbeat": {
    "enabled": true,
    "interval": 30
  }
}

• enabled: Enable/disable heartbeat
• interval: Minutes between heartbeat checks (minimum 5)

Device Monitoring

Monitor USB and hardware devices (Linux only):

{
  "devices": {
    "enabled": false,
    "monitor_usb": true
  }
}

Complete Example

Here’s a production-ready configuration:

{
  "agents": {
    "defaults": {
      "workspace": "~/.weaver/workspace",
      "restrict_to_workspace": true,
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "max_tokens": 16384,
      "temperature": 0.7,
      "max_tool_iterations": 20
    }
  },
  "providers": {
    "anthropic": {
      "api_key": "${ANTHROPIC_API_KEY}"
    },
    "openai": {
      "api_key": "${OPENAI_API_KEY}"
    }
  },
  "channels": {
    "telegram": {
      "enabled": true,
      "token": "${TELEGRAM_BOT_TOKEN}",
      "allow_from": ["123456789"]
    },
    "discord": {
      "enabled": true,
      "token": "${DISCORD_BOT_TOKEN}",
      "allow_from": ["987654321"]
    }
  },
  "tools": {
    "web": {
      "brave": {
        "enabled": true,
        "api_key": "${BRAVE_SEARCH_API_KEY}",
        "max_results": 5
      },
      "duckduckgo": {
        "enabled": true,
        "max_results": 5
      }
    }
  },
  "heartbeat": {
    "enabled": true,
    "interval": 30
  },
  "gateway": {
    "host": "0.0.0.0",
    "port": 18790
  }
}

Configuration Priority

Weaver loads configuration in this order (later sources override earlier ones):

1. Default configuration (pkg/config/config.go:218)
2. ~/.weaver/config.json
3. Environment variables

This allows you to commit config.json to version control while keeping secrets in environment variables.