Follow these steps to download the dataset used for SFI computation and vector database creation:
- European City Data:
- Download european-city-data dataset
- Save under
backend/european-city-datadirectory
- Tripadvisor Data:
- Download european-city-tripadvisor-data dataset
- Save under
backend/european-city-data/data-sources/tripadvisordirectory
The system has several configuration options that can be set via the config.ini file in the backend directory.
- Note: For sections "retrieval" and "indexing" certain configurations are pre-configured in the backend and don't require manual adjustment.
Configure the IP address and port for hosting the FastAPI server.
Configure the necessary parameters for tasks that use pre-trained LLMs.
Configure the retriever, reranking, and context limit.
Configure the file path used for saving the logs.
Set the following environment variables in backend/.env. VertexAI authentication additionally requires a JSON file with the credentials under the path backend/.config/gcp_default_credentials.json.
| Variable | Description |
|---|---|
| DATA_REPO | Path of the repository or file where the data is stored |
| HF_TOKEN | Hugging Face API Token |
| OPENAI_API_KEY | OpenAI API Key |
| VERTEXAI_PROJECTID | Vertex AI Project ID |
| RAGAS_APP_TOKEN | RAGAS API token for uploading/visualizing evaluation results |
Install Pandoc for Wikitext to Markdown conversion.
MacOS (Homebrew)
brew install pandocLinux (apt)
sudo apt-get install pandoccd backend
python3.12 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python3 src/vectordb/create_db.py
python3 server.pyThis project incorporates inspired/modified code from:
- llm-convrec (MIT License) - see THIRD-PARTY-LICENSES.txt for license details.
- green-city-finder
