LLM-Driven Restaurant Finder API

This project is a FastAPI-based service that leverages OpenAI and Foursquare Places API to find restaurants based on natural language queries.
The application is containerized with Docker, deployed via Terraform (IaC), and integrated with GitHub Actions for CI/CD.

🚀 Features

• Healthcheck

  • GET /api/ping – Simple ping endpoint to verify that the API is running.

• Restaurant Finder

  • GET /api/execute

    • Accepts a natural language message (e.g., “Find me Japanese restaurants near Manila with good ratings”).

    • Calls OpenAI to parse the message into a structured JSON request accepted by Foursquare Places API.

    • Returns a standardized Restaurant DTO:

      class RestaurantDto(BaseModel):
          name: str
          cuisine: str
          address: str
          operating_hours: str | None = None
          rating: float | None = None
          price_level: int | None = None

• Infrastructure

  • Terraform for infrastructure as code (/infra).
  • GitHub Actions for automated builds & pushing Docker images to ECR.
  • Optional CloudWatch support.

• Domain

  • Configured with DuckDNS for custom domain setup.

🛠️ Local Development

Requirements

• Python 3.11+
• Poetry for dependency management
• Docker (for containerized builds)
• Terraform (if you want to deploy infra)
• OpenAI and Foursquare API keys (for /api/execute)

Setup

# Install dependencies
poetry install

# Activate virtual environment
$(poetry env activate)

# Run the app
python -m app

The API will be available at http://localhost:8000. Interactive Swagger docs: http://localhost:8000/docs

🧪 Testing

The project uses pytest. Note that you’ll need valid API keys for OpenAI and Foursquare:

OPENAI_API_KEY=your_key \
FOURSQUARE_SERVICE_KEY=your_key \
pytest -sv

📦 Docker

Build and run with Docker:

# Build image
docker build -t restaurant-finder-api .

# Run container
docker run --env-file .env -p 8000:8000 restaurant-finder-api

☁️ Infrastructure

Terraform is used to provision infra in AWS:

cd infra
terraform init
terraform apply

⚠️ Don’t forget to build and push the Docker image to AWS ECR before applying Terraform.

🔑 Environment Variables

Variable	Description	Required
OPENAI_API_KEY	OpenAI API key	✅
FOURSQUARE_SERVICE_KEY	Foursquare API key	✅
CW_LOG_GROUP_NAME	CloudWatch Log Group name	❌
CW_NAMESPACE	CloudWatch namespace for custom metrics	❌
INSTANCE_ID	EC2/ECS instance identifier for metrics	❌

🌐 Domain

This project uses DuckDNS for domain management. Update DNS accordingly after deployment.

📋 Roadmap / Improvements

• Save requests/responses (parsing) from OpenAI for dataset
• Add caching layer for frequent queries
• Improve test coverage (mock OpenAI & Foursquare calls)
• Expand infra to support multi-region deployments
• Add JWT authentication/authorization
• Expand SearchPlacesDto to accomodate other query