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.
- 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.
- Supports location-aware querying over water resource datasets.
- Retrieves contextual information about lakes, reservoirs, and other water bodies.
- Numerical and ranking-based questions are answered using Pandas-driven data processing.
- Ensures accurate results for queries involving area, depth, capacity, and comparative 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.
- Native Speech-to-Text and Text-to-Speech integration using browser APIs.
- Enables natural conversational access to geographic information.
- Fuzzy matching for geographic entities and location names.
- Improves retrieval quality by handling spelling variations and user input errors.
-
User submits a text or voice query.
-
FastAPI backend processes and classifies the request.
-
Query router selects the appropriate pipeline:
- Semantic Retrieval (FAISS)
- Structured Geographic Data Retrieval
- Deterministic Analytics Engine
- Engineering Suitability Engine
-
Retrieved information is combined into a grounded context.
-
Gemini generates a final response based on verified data.
-
Results are displayed in the React dashboard and optionally delivered through voice output.
- React.js
- Vite
- Native Web Speech API
- FastAPI
- Python
- Uvicorn
- Google Gemini
- FAISS Vector Database
- Text Embeddings
- Structured Geographic Knowledge Base
- Pandas
- 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
- Multi-region geospatial support
- Real-time GIS map integration
- Satellite imagery analysis
- Advanced engineering recommendation models
- Multi-agent planning workflows
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.