Skip to content

Repository files navigation

GeoRAG

An AI-powered geospatial decision support system built for water resource planning and management. The platform combines Retrieval-Augmented Generation (RAG), geographic information systems (GIS), and deterministic analytics to provide accurate insights about water bodies, infrastructure suitability, and location-specific queries.

Designed for government and planning use cases, the system minimizes hallucinations by combining semantic retrieval with structured geographic data and rule-based reasoning.

Features

Hybrid Retrieval Architecture

  • Combines FAISS vector search, structured geographic datasets, and rule-based reasoning.
  • Routes queries through the most suitable retrieval pipeline instead of relying solely on LLM generation.

Geospatial Intelligence

  • Supports location-aware querying over water resource datasets.
  • Retrieves contextual information about lakes, reservoirs, and other water bodies.

Deterministic Analytics

  • Numerical and ranking-based questions are answered using Pandas-driven data processing.
  • Ensures accurate results for queries involving area, depth, capacity, and comparative analysis.

Engineering Suitability Analysis

  • Evaluates the suitability of water bodies for infrastructure such as dams and anicuts using predefined engineering constraints.
  • Combines geographic attributes with domain-specific rules.

Voice-Enabled Interaction

  • Native Speech-to-Text and Text-to-Speech integration using browser APIs.
  • Enables natural conversational access to geographic information.

Intelligent Query Handling

  • Fuzzy matching for geographic entities and location names.
  • Improves retrieval quality by handling spelling variations and user input errors.

System Architecture

  1. User submits a text or voice query.

  2. FastAPI backend processes and classifies the request.

  3. Query router selects the appropriate pipeline:

    • Semantic Retrieval (FAISS)
    • Structured Geographic Data Retrieval
    • Deterministic Analytics Engine
    • Engineering Suitability Engine
  4. Retrieved information is combined into a grounded context.

  5. Gemini generates a final response based on verified data.

  6. Results are displayed in the React dashboard and optionally delivered through voice output.


Tech Stack

Frontend

  • React.js
  • Vite
  • Native Web Speech API

Backend

  • FastAPI
  • Python
  • Uvicorn

AI & Data Layer

  • Google Gemini
  • FAISS Vector Database
  • Text Embeddings
  • Structured Geographic Knowledge Base
  • Pandas

Key Learnings

  • Retrieval-Augmented Generation (RAG)
  • Vector databases and semantic search
  • Geospatial data processing
  • FastAPI backend development
  • AI system reliability and hallucination reduction
  • Hybrid retrieval architectures
  • Rule-based reasoning systems

Future Improvements

  • Multi-region geospatial support
  • Real-time GIS map integration
  • Satellite imagery analysis
  • Advanced engineering recommendation models
  • Multi-agent planning workflows

Use Case

This platform was developed as part of a geospatial AI initiative focused on supporting water resource planning and decision-making through reliable, retrieval-grounded AI systems.

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages