Skip to content

README.md

Latest commit

 

History

History
258 lines (167 loc) · 5.92 KB

README.md

File metadata and controls

258 lines (167 loc) · 5.92 KB
Raw

Step 1: Ollama & Clawdbot Setup

This guide will help you install and configure Ollama and Clawdbot, the foundation of your local AI system.

Overview

Ollama is a lightweight framework for running large language models locally, while Clawdbot is an MCP (Model Context Protocol) client that acts as the orchestration layer, connecting your AI model to various tools and services.

Prerequisites

  • Operating System: Windows, macOS, or Linux
  • RAM: 8GB minimum (16GB+ recommended)
  • Disk Space: 10-50GB depending on model choice
  • Node.js: Version 18+ (Download here)

Part 1: Installing Ollama

For Windows

  1. Download the Ollama installer from ollama.ai/download
  2. Run the installer and follow the installation wizard
  3. Verify installation by opening PowerShell and running: 
    ollama --version

For macOS

  1. Download from ollama.ai/download or use Homebrew: 
    brew install ollama
  2. Verify installation: 
    ollama --version

For Linux

curl -fsSL https://ollama.ai/install.sh | sh
ollama --version

Part 2: Downloading AI Models

Clawdbot requires models with tool calling capabilities. Recommended models for 2026:

Recommended: Qwen3-Coder

Best balance of performance and resource usage for tool calling:

ollama pull qwen3-coder

Alternative: Mistral-Small

Good for lighter systems:

ollama pull mistral-small

High Performance: GLM-4.7-Flash

For powerful systems:

ollama pull glm-4.7-flash

Test Your Model

ollama run qwen3-coder

Type a test message like "Hello, can you help me?" and press Enter. If you get a response, your model is working. Type /bye to exit.

Part 3: Installing Clawdbot

Install Clawdbot

Clawdbot is typically installed via npm or a dedicated installer. Check the official Clawdbot repository for the latest installation method:

# Example installation (verify with official docs)
npm install -g clawdbot

# Or if using a dedicated installer, download from the official site

Initial Configuration

  1. Run the onboarding wizard:

    clawdbot onboard

  2. Configure Ollama as your LLM provider:

    During onboarding, select:

    • Provider: Ollama
    • Model: qwen3-coder (or your chosen model)
    • Base URL: http://localhost:11434

  3. Alternatively, edit config manually:

    Open your Clawdbot configuration file:

    • Windows: %APPDATA%\clawdbot\config.yaml
    • macOS/Linux: ~/.config/clawdbot/config.yaml

    Add or edit:

    llm:
  provider: ollama
  model: qwen3-coder
  base_url: "http://localhost:11434"
  num_ctx: 32768 # Context window size

Part 4: Starting the Clawdbot Gateway

The gateway is the background service that keeps Clawdbot running:

# Start the gateway
clawdbot gateway start --daemon

# Check status
clawdbot gateway status

# View logs
clawdbot gateway logs

Part 5: Accessing the Dashboard

Clawdbot includes a web dashboard for management:

  1. Open your browser and navigate to: http://localhost:18789
  2. You should see the Clawdbot Control UI
  3. Verify that:
    • Ollama connection shows as Connected
    • Your chosen model is listed
    • Gateway status is Running

Part 6: Testing the Setup

Test via Command Line

clawdbot chat

This opens an interactive chat session. Try asking:

  • "What is 2+2?"
  • "Can you help me with coding?"

Type exit to close.

Test Context Window

For web search and browser automation to work properly, ensure your model has sufficient context:

ollama run qwen3-coder --num_ctx 32768

Test with a longer prompt to verify it handles extended context.

Important Configuration Notes

Context Window Size

Web searching returns large amounts of data. Ensure your model configuration includes adequate context:

llm:
  num_ctx: 32768 # Minimum 32k for reliable web search

Performance Tuning

If responses are slow:

  • Close unnecessary applications to free up RAM
  • Consider using a smaller model like mistral-small
  • Adjust num_ctx to balance performance and capability

Persistence

To ensure Clawdbot starts automatically on system boot:

Windows:

# Create a startup task
clawdbot gateway install-service

macOS/Linux:

# Create a systemd service or launchd plist
clawdbot gateway install-service

Troubleshooting

Ollama Not Found

  • Verify Ollama is in your PATH
  • Restart your terminal after installation

Connection Refused

  • Ensure Ollama is running: ollama serve
  • Check if another service is using port 11434

Model Loading Errors

  • Verify model is downloaded: ollama list
  • Re-pull the model: ollama pull qwen3-coder

Gateway Won't Start

  • Check for port conflicts on 18789
  • Review logs: clawdbot gateway logs

Security Best Practices

  1. Dashboard Access: By default, the dashboard is only accessible on localhost. Keep it this way until you set up Tailscale (Step 4).

  2. API Keys: Never commit your config.yaml to version control. Add it to .gitignore.

  3. User Permissions: Don't run Clawdbot as administrator/root. Use a standard user account.

Next Steps

Once your Ollama and Clawdbot setup is complete and the dashboard shows everything connected:

➡️ Continue to Step 2: Brave MCP Server Setup to add internet search capabilities.

Additional Resources

← Back to Main README | Next: Brave Search MCP →