Grounded insights, with citations you can trust
A research workspace for reading source material, exploring citations, and drafting answers with human review.
No single tool retrieves, reasons, and cites.
Knowledge workers drown in unstructured information across disconnected tools, and the usual options each fall short in a different way.
- Search
- Returns keywords, not understanding. You still read everything.
- LLM chat
- Reasons fluently, then hallucinates without a source you can check.
- Note apps
- Store what you write. They never connect or reason across it.
NOUS is one grounded workspace that does all three, and shows its work.
What NOUS does
A research-grade pipeline, from ingest to a cited draft you can export.
Agent system, with a human in the loop
Keep a human in control of every write with an intent-routed LangGraph StateGraph for research, writing, and data. Destructive tools pause for review, and Postgres checkpoints keep threads alive across restarts.
Hybrid retrieval
Find answers across text, images, audio, and citations with keyword search and graph traversal fused and reranked through Postgres FTS, Neo4j, and Cohere.
Citation pipeline
Trace every claim back to a source through extraction from arXiv, Semantic Scholar, CrossRef, and PDFs. Export the resulting citation graph to BibTeX, APA, IEEE, or MLA.
Multi-agent literature reviews
Turn a literature review into a versioned draft with theme extraction, synthesis, inline citations, and LaTeX/.bib export.
Multimodal ingest
Bring PDFs, images, audio, and video into one index with OCR, scene detection, Whisper transcription, diarization, and frame extraction.
Local AI in the browser
Run privacy-sensitive workflows in the browser with WebLLM models including Llama, Phi, Gemma, and Qwen. No data leaves the client.
Trust is measured, not claimed
These are internal DeepEval CI thresholds, not production performance claims. A regression past any threshold blocks the merge, so grounding does not drift over time.
- >90%
- Faithfulness threshold
- internal grounded evals
- >70%
- Answer relevancy threshold
- internal grounded evals
- <10%
- Hallucination threshold
- CI regression gate
One graph, end to end
A checkpointed StateGraph routes intent to the right subgraph, runs tools, and saves memory. The interrupt and resume cycle keeps a human in control of every write.
- rag
- intent
- memory
- route
- tools
- save
Built on
- Next.js 15
- FastAPI
- LangGraph
- Neo4j
- Postgres
- Redis
- Cohere rerank
- Whisper
Read, reason, and write with sources you can trust
Explore the live showcase, or read the source. NOUS is open and built in the open.