Joshua Sundance Bailey
readme
53e0346
|
raw
history blame
3.38 kB
metadata
title: govgis_nov2023-slim-faiss
emoji: 🌎
colorFrom: green
colorTo: blue
sdk: streamlit
sdk_version: 1.28.2
app_file: app.py
pinned: true
license: mit

govgis_nov2023-slim-faiss

License: MIT python

Push to HuggingFace Space Open HuggingFace Space

pre-commit Ruff Checked with mypy Code style: black

security: bandit

govgis_nov2023-slim-faiss

πŸ€– This README was written by GPT-4. πŸ€–

Features

  • Semantic Search on GIS Metadata: Leverages the govgis_nov2023 dataset to provide detailed insights into numerous GIS servers and layers.
  • Natural Language Query Processing: Uses Claude-Instant and Claude-2.1 models to interpret and rephrase user queries (optional).
  • Advanced Document Retrieval: Integrates FAISS vector store for efficient and relevant document retrieval based on query semantics.
  • Customizable User Experience: Sidebar controls to adjust search parameters and input fields for queries.

Dataset Overview

  • Content: The app is built around the govgis_nov2023 dataset, which documents metadata from 1684 government ArcGIS servers, detailing almost a million individual layers.
  • Unique Snapshot: Provides a unique snapshot of these servers, with metadata including field information for feature layers and cell size for raster layers.

User Interface Guide

  • Adjust search settings like result limits and response generation parameters in the sidebar.
  • Securely enter your Anthropic API key for model access.
  • Submit natural language queries related to GIS data.

Contributions

We welcome contributions. Please follow the standard fork and pull request process.

Support and Contact

For support, please raise an issue on GitHub or in the HuggingFace space.

License

This project is under the MIT License.

Acknowledgments

Thanks to the Huggingface and Streamlit communities, and special acknowledgment to Joseph Elfelt and the creators of the restgdf library for their contributions to the GIS field.

TODO

  • Add an open source model like HuggingFaceH4/zephyr-7b-beta
  • Hybrid search w/ bm25 or similar
  • Find a lightweight way to incorporate geospatial filtering
  • Add more parameters