ollama-coder 0.1.2

Agentic coding assistant for Ollama

OllamaCode

Agentic coding assistant for Ollama with tool use, autonomous mode, and project/user configuration.

Install

From PyPI (after you publish):

pip install ollama-coder

For local development:

pip install -e .

Run

ollama-coder

You will be prompted to choose a model on startup. The selected model can be saved as your default.

Common Commands

• /auto toggle autonomous mode
• /models list installed models
• /model show or set the active model
• /host show or set the Ollama host (for remote/cloud)
• /config show effective configuration
• /clear clear conversation history
• /quit exit

Configuration

User config lives at ~/.ollamacode/settings.json. Example:

{
  "model": "llama3:latest",
  "max_iterations": 25,
  "max_tool_rounds": 8,
  "ollama": {
    "host": "http://127.0.0.1:11434",
    "timeout_sec": 60,
    "headers": {},
    "api_key": ""
  },
  "web_search": {
    "enabled": false,
    "provider": "custom",
    "endpoint": "",
    "api_key": "",
    "timeout_sec": 15,
    "max_results": 5
  }
}

You can also set OLLAMA_HOST in your environment to point at a remote server. The /host command updates the config and rebuilds the client at runtime.

Requirements

• Ollama server running locally or reachable via OLLAMA_HOST
• Python 3.9+

GitHub

Repository: https://github.com/lalomorales22/ollama-code

License

MIT

Publish to PyPI

Manual publish:

python -m build
python -m twine upload dist/*

Auto Publish (GitHub Actions)

This repo includes a GitHub Actions workflow that publishes to PyPI on tags.

Steps:

1. Bump the version in pyproject.toml and ollama_code/__init__.py
2. Commit the change
3. Tag and push:

   git tag v0.1.1
   git push origin v0.1.1
   

Trusted Publisher setup (one-time):

• In PyPI, add a Trusted Publisher for this repo.
• Workflow name (file): publish.yml (or .github/workflows/publish.yml)
• Environment name: leave blank (unless you add one)