Hybrid RAG · Gemini 2.0 · pgvector
Noguessing.
Nohallucination.

Upload documents, ask questions, get answers grounded in your actual content — with exact citations showing document, page, and passage.

Zero hallucinationExact citations< 3s response
Scroll
Powered by
Gemini 2.0 Flashpgvector HNSWBM25 Full-textCross-encoder RerankingRedis CacheCelery AsyncNext.js 16Django 5FastAPI
Why VeritasRAG

Why users stick with it.

Every layer is engineered to maximize retrieval quality and response grounding. Zero hallucination is a design constraint, not a goal.

Hybrid Retrieval

pgvector HNSW cosine ANN + BM25 full-text fused via reciprocal rank fusion. Maximum precision at every query.

20→5candidates reranked

Grounded Answers

Every response is anchored to exact source passages. A grounding score (0–1) flags low-confidence answers automatically.

0.94avg similarity score

Sub-3s Responses

Repeated queries resolve from Redis cache in under 5ms. Cold RAG completes end-to-end in under 3 seconds.

< 5mscache hit latency

Exact Citations

Chunk text, source document, page number, similarity score — every answer fully traceable back to the source.

100%answers cited
Process

The Process. Fast, Clear, Done.

No manual work. No waiting. Just answers grounded in your content.

01

Upload Your Docs

PDF, DOCX, or TXT. Direct Cloudinary upload. Async ingestion via Celery — the UI never blocks while your document processes.

02

Ask in Plain Language

Type any question about your documents. Hybrid search finds the most relevant passages across all selected files instantly.

03

Get Cited Answers

Grounded answers with exact citations: document, page, chunk, similarity score. Low confidence? A warning tells you.

FAQ

Questions? We've got answers.

We've answered the most common questions below. If you need more, just ask.

Get Started

Ready to ask better questions?

Upload your first document and get grounded, cited answers in under 3 seconds. Free to start, no credit card required.

Get Started Now