This project enables converting natural language questions into executable SQL queries using a combination of semantic embeddings, vector search, and prompt-based generation. Built with Spring Boot and integrated with Hugging Face APIs and PostgreSQL (pgvector), it provides a scalable backend for text-to-SQL translation.
- 馃挰 Accepts user input in plain English and returns relevant SQL queries
- 馃 Uses Hugging Face for embedding generation and query generation
- 馃搳 Performs vector similarity search over schema and metadata stored in pgvector
- 馃攳 Dynamically builds context from metadata to guide query generation
鈿狅笍 Returns fallback response when schema context is insufficient for query formation
- Backend: Java + Spring Boot
- Embeddings & Chat: Hugging Face APIs
- Database: PostgreSQL with
pgvectorextension - Search Method: Hybrid (semantic + metadata relevance)
- API Style: REST