QueryMind is a full-stack AI application that translates natural language into validated, executable SQL — and runs it against your PostgreSQL database in real time.

You connect a database. You type a question. QueryMind retrieves the relevant schema from Pinecone, generates precise SQL using Gemini 2.5 Flash via a streaming SSE pipeline, validates it for safety, executes it on your Neon DB, and returns the results — all in under two seconds.

It also ships a Schema Designer: describe your database in plain English and get a full ER diagram on an interactive React Flow canvas, with auto-drawn foreign key relationships and PDF/SQL export.

No BI tool setup. No SQL editor. No developer in the loop.

PIPELINE ARCHITECTURE:-

User question

│

▼

Gemini text-embedding-004

(embed the question)

│

▼

Pinecone vector search

(top-6 relevant table docs by cosine similarity)

│

▼

Gemini 2.5 Flash

(schema context + question → SQL, streamed via SSE)

│

▼

SQL Validator

(keyword blocklist + sqlparse + table existence check)

│

▼

Query Executor

(read-only async connection → Neon DB → 500-row cap)

│

▼

Results → Frontend

(streamed, rendered, exportable as CSV)