Mino

Local PDF Q&A: upload documents, ask questions, and chat with your files. Everything runs in Docker — no separate Ollama install required.

Requirements

• Docker Desktop (macOS / Windows) or Docker Engine + Compose (Linux)
• ~8 GB RAM recommended (AI models + PDF indexing)
• Several GB free disk space for downloaded models

Optional: change models before installing

Most users can skip this. Defaults work out of the box:

• Chat model: gemma4
• Embed model: nomic-embed-text

To use different Ollama models, do this before your first launch:

cp .env.example .env
# Edit .env and set CHAT_MODEL / EMBED_MODEL

On first install, Mino shows which models it will download and pauses so you can stop and edit .env if needed.

Quick start

git clone <your-repo-url> mino
cd mino
./scripts/launch.sh

That's it. The first run downloads Docker images and AI models (this can take 10+ minutes). Later launches start in seconds and open your browser automatically.

macOS: you can also double-click Launch Mino.command.

First run

When you launch for the first time, you'll see a setup banner listing the chat and embed models. Press Enter to continue. Mino will:

1. Build the app containers
2. Start Ollama inside Docker
3. Download the configured models
4. Start all services and open http://localhost:8000

Stop Mino

./scripts/stop.sh

Your indexed documents and chat history are kept unless you run docker compose down -v.

Troubleshooting

Docker is not running
Start Docker Desktop and run ./scripts/launch.sh again.

Port 8000 is already in use
Stop the other service using that port, or change the published port in docker-compose.yml.

Model download fails or chat errors mention a missing model
Check the model name exists in Ollama's library. Update .env, remove .mino-installed, and run ./scripts/install.sh again.

Windows without bash
Use Git Bash or WSL to run the scripts, or manually:

docker compose up -d --build

Then open http://localhost:8000 in your browser.

Developers

Start the stack directly:

docker compose up --build

Run tests (no Docker required):

python -m unittest api/tests/test_retrieval.py
python -m unittest librarian/tests/test_parsing.py

Environment variables (OLLAMA_HOST, REDIS_HOST, CHAT_MODEL, EMBED_MODEL) are set in docker-compose.yml. See .env.example for the user-facing overrides.