ValGrace/travel_assistant

RAG pipeline for tourism in Kenya (find accommodations, hidden gems, plan a trip, compare prices, plan a safari). # Kenyan Tourism Intelligence App This project is a small travel assistant for Kenya built with FastAPI, LangChain, Chroma, and Gemini. It loads local tourism data, embeds it for retrieval, and answers trip-planning questions such as destinations, hotels, campsites, parks, and safari itineraries. ## Project purpose The app lets a user ask natural-language questions like: - "Plan a 3-day safari in Maasai Mara" - "Suggest hotels in Nairobi under a certain budget" - "What are the best parks and lodges near Mombasa?" It uses a vector database and reranking pipeline to find relevant content and then generates a structured travel response. ## Architecture ## Setup 1. Clone the repository and open it in your terminal. 2. Install `uv` if it is not already available: ```bash pip install uv ``` 3. Create and activate the project environment with uv: ```bash uv venv source .venv/bin/activate ``` Windows PowerShell: ```powershell uv venv .\.venv\Scripts\Activate.ps1 ``` 4. Install dependencies: ```bash uv pip install -e . ``` 5. Create a `.env` file in the project root with your Gemini API key: ```env GEMINI_API_KEY=your_api_key_here ``` 5. Confirm the local vector database exists in `data/chroma_db`. ## Run the app Start the API with: ```bash uv run main.py ``` The app will be available at: - 127.0.0.1 - 127.0.0.1 ## Test it from the command line ### Health check ```bash curl 127.0.0.1 ``` Expected output: ```json {"status": "App is running"} ``` ### Query the travel assistant ```bash curl -X POST 127.0.0.1 \ -H "Content-Type: application/json" \ -d '{"question":"Plan a 3-day safari in Maasai Mara with lodge options and budget guidance."}' ``` This should return a JSON response with fields such as `intro`, `highlights`, `itinerary`, `missing_info`, and `follow_up_questions`. ## Notes - The first startup may download embedding and reranking models, so network access may be required. …