Licensed under the MIT License. See LICENSE for details.

OwA — Ollama Workspace Agent

An open-source local coding assistant powered by Ollama. Runs entirely on your own hardware — no cloud, no telemetry, no API keys required beyond your own Ollama server.

Install

pip install ollama-workspace-agent

Or with pipx for an isolated install:

pipx install ollama-workspace-agent

Requirements

• Python 3.11+
• Ollama running on a reachable machine
• A chat model installed on Ollama, such as llama3.1:8b
• An embedding model installed on Ollama, such as nomic-embed-text

Quick Start

cd your-project
owa

On first run, type /setup to configure your Ollama connection. OwA will walk you through it interactively and save the config to ~/.config/owa/.env.

Then just start chatting — OwA will auto-index your project on startup.

Configuration

Run /setup inside OwA, or manually create ~/.config/owa/.env:

LLM_BASE_URL=http://YOUR_OLLAMA_HOST:11434/v1
LLM_MODEL=llama3.1:8b
EMBEDDING_BASE_URL=http://YOUR_OLLAMA_HOST:11434
EMBEDDING_MODEL=nomic-embed-text
API_KEY=choose-a-private-api-key

You can also place a .env in your project root to override the global config for that project.

LLM_BASE_URL uses Ollama's OpenAI-compatible /v1 API. EMBEDDING_BASE_URL uses Ollama's native /api/embed API. If Ollama runs on the same machine, replace YOUR_OLLAMA_HOST with 127.0.0.1.

Verify Ollama

curl http://YOUR_OLLAMA_HOST:11434/api/tags

A successful response contains a models list. Make sure the models in your config are installed:

ollama pull llama3.1:8b
ollama pull nomic-embed-text

CLI Commands

API Mode

Run the HTTP API:

uvicorn app.api:app --host 127.0.0.1 --port 8000

Use the CLI as an API client:

owa --api-url http://127.0.0.1:8000

Available endpoints: GET /health, GET /status, POST /chat, POST /chat/stream, POST /clear, POST /index.

When API_KEY is configured, include it in the X-API-Key header for every endpoint except /health.

Tools

The agent can call these workspace tools:

Project Layout

.
├── app/
│   ├── agent/       Agent orchestration and system prompt
│   ├── api.py       FastAPI endpoints and request models
│   ├── cli.py       CLI entry point
│   ├── config.py    Global config path resolution
│   ├── service.py   Shared chat, indexing, and status service
│   ├── indexer/     Code indexing and semantic search
│   ├── llm/         OpenAI-compatible Ollama client
│   └── tools/       Workspace tool implementations and registry
├── demo/            Small demo code
├── main.py          Thin shim for `python main.py`
├── CHANGELOG.md     Project change history
└── README.md        Project documentation

Development

Clone and install in editable mode:

git clone https://github.com/ZakaCoding/ollama-workspace-agent
cd ollama-workspace-agent
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Run tests:

pytest tests/ -v

Compile check:

python -m py_compile main.py app/agent/*.py app/indexer/*.py app/llm/*.py app/tools/*.py