An AI-powered restaurant finder built with open-source components:

• FastAPI + LangGraph for orchestration
• Chainlit for chat UI
• SearchAPI for restaurant data retrieval
• SQLite for conversation memory
• Docker Compose for deployment
• OpenTelemetry + Jaeger + Prometheus + Grafana for observability

.
├── restaurant-finder-api/
│   ├── src/
│   │   ├── application/orchestrator/
│   │   ├── domain/
│   │   ├── evaluation/
│   │   └── infrastructure/
│   ├── tests/
│   ├── Dockerfile
│   ├── Makefile
│   └── pyproject.toml
├── restaurant-finder-ui/
│   ├── app.py
│   ├── .env.example
│   └── pyproject.toml
└── restaurant-finder-infra/
    ├── docker-compose.yml
    ├── prometheus/prometheus.yml
    └── grafana/provisioning/

• Python 3.11+
• uv
• Docker + Docker Compose
• Search API key
• OpenAI-compatible model API key (default GROQ_API_KEY)

1. Configure environment files:

cp restaurant-finder-api/.env.example restaurant-finder-api/.env
cp restaurant-finder-ui/.env.example restaurant-finder-ui/.env

2. Set required values in restaurant-finder-api/.env:

• GROQ_API_KEY
• SEARCH_API_KEY

3. Start full stack:

cd restaurant-finder-infra
docker compose up --build

4. Open services:

• UI: http://localhost:8000
• API health: http://localhost:8080/health
• Jaeger: http://localhost:16686
• Prometheus: http://localhost:9090
• Grafana: http://localhost:3000

cd restaurant-finder-api
uv sync
uv run python -m src.main

API runs on http://localhost:8080.

cd restaurant-finder-ui
uv sync
chainlit run app.py

In the UI, click View Memory to inspect saved preferences, facts, and summaries for the current customer_name + conversation_id in a side panel.

• GET /memory
• Query params:
  • customer_name (required)
  • conversation_id (required)
  • limit_per_category (optional, default 10, max 50)
• Returns category buckets and totals for preferences, facts, and summaries across current-session + same-user scope.

The evaluation framework is OSS/local and can run against the local API endpoint.

cd restaurant-finder-api
make eval
make eval-categories CATEGORIES="basic_search dietary_search"
make eval-safety

• deploy-image.yml: runs API tests, then builds and publishes API image to GHCR
• deploy-infra.yml: validates Docker Compose and API image build
• destroy-infra.yml: teardown guidance for local docker stack

cd restaurant-finder-infra
docker compose down -v