Our Enterprise Tech Stack
Optimized for low latency ingestion, vector search performance, and secure data isolation.
Vector Database
PostgreSQL & pgvector
Stores document chunk embeddings and metadata indices. Uses HNSW index methods for rapid similarity queries.
Backend API Service
FastAPI & Uvicorn
Asynchronous endpoint handling, file upload management, RAG pipelining, and user authentication with JWT tokens.
Frontend Interface
Next.js 15 & Tailwind CSS
Modern page routing, stateful dashboard views, custom styled tailwind utilities, and responsive design systems.
Task Queues
Celery & Redis Broker
Handles resource-heavy processes like PDF document chunking and vector index embedding in the background.
Hosting & Proxy
Docker & Nginx
Containerized deployment, SSL encryption, rate limiting, and reverse proxying mapping all traffic routes.
RAG architecture for cited business answers
ChatWithDB separates ingestion, embedding, retrieval, and response generation so each answer can stay fast and auditable. PostgreSQL pgvector handles semantic search, FastAPI coordinates document and database context, and the Next.js interface keeps citations visible to the user.
Retrieval flow
- 1. Parse: PDFs, CSV files, Excel sheets, text files, and records are normalized into searchable chunks.
- 2. Embed: Chunks are stored with metadata and source references in PostgreSQL pgvector.
- 3. Retrieve: Natural language questions search semantic and structured context together.
- 4. Cite: Answers return document names, row context, and supporting excerpts.
Experience Our Architecture Live
Fast, secure, and production-ready. Create your workspace now and start querying with zero config required.