RecipeRAG

Recipe assistant built with Retrieval-Augmented Generation, FastAPI, and Docker.

⚙️ Installation

🔧 Set Up the Python Environment

1. Clone the repository

2. Install uv — A fast Python package manager
   📖 Installation guide

3. Create and activate a virtual environment

   uv venv .venv
   source .venv/bin/activate

4. Install dependencies

   Development setup (app + dev tools):

   uv sync

   Production-only (skip dev tools):

   uv sync --no-group dev

5. Enable pre-commit hooks [Optional]

   uv run pre-commit install

🚀 Running the App

🔹 Option A --- Run everything with Docker Compose (recommended)

All services (API, Postgres + pgvector, Ollama) can be started in one command:

docker compose up

• API will be available at: http://localhost:8000
• Ollama runs on port 11434
• Postgres runs on port 5432

🔹 Option B --- Run with Taskfile (local dev)

If you prefer to manage services separately:

1. Start Ollama

   task ollama:up

2. Start Postgres

   task db:up

3. Run the FastAPI app

   task app:run

📄 Example .env

Create a .env file in the project root with your environment variables:

OPENAI_API_KEY=ollama
OPENAI_BASE_URL=http://ollama:11434/v1
EMBEDDING_MODEL=nomic-embed-text

POSTGRES_USER=postgres
POSTGRES_PASSWORD=password
POSTGRES_HOST=db
POSTGRES_DB=mydb
POSTGRES_PORT=5432

DATABASE_URL="postgresql+psycopg://${POSTGRES_USER}:${POSTGRES_PASSWORD}@${POSTGRES_HOST}:${POSTGRES_PORT}/${POSTGRES_DB}"

⚠️ Inside Docker Compose, use POSTGRES_HOST=db (the service name), not localhost.

📓 Usage Example (Notebook)

• Without RAG - see the existing notebook: notebooks/demo_without_rag.ipynb

Walks through creating an assistant, creating a chat, and getting completions.

• With RAG (ingredients + directions injected) - see the new notebook: notebooks/demo_with_rag.ipynb

Shows the same flow, but calls the /completions endpoint with ?rag=true.

✅ Prerequisite for RAG: run the offline indexer to populate the vector store

task app:index