Answers with sources, straight to the map
Ask in plain language. VOL AI retrieves the evidence, writes a grounded answer, cites every claim and takes you to the cluster where it lives.
Trusted by fast-growing research teams around the world
- Helix Labs
- northwind
- Quanta
- atlas.bio
- Meridian
- COBALT
- Lumen Institute
- orbital
From question to evidence in under two seconds
A retrieval-augmented pipeline tuned for research corpora.
- 01RetrieveHybrid semantic and keyword search over the corpus you chose.
- 02RerankA reranker keeps only the passages that really answer the question.
- 03AnswerA concise answer is written strictly from those passages.
- 04Cite & flyEach claim is cited; clicking it flies the map to the node and its cluster.
An assistant built for research
Cluster summaries
Answers group sources into themed clusters you can open on the map.Follow-up questions
Conversations keep context across turns.Multilingual
Ask in 40+ languages, across sources in any language.Streaming
Answers stream token by token with citations at the end.Grounded only
If the corpus does not know, VOL AI says so instead of guessing.- median answer
- 1.6s
- answers with sources
- 98%
- languages
- 40+
- data used for training
- 0
Context around every answer
Answers are never a dead end. Every citation is a node; every group of citations is a cluster you can open, explore and share.
- Highlight cited nodes
- Open RAPTOR-style answer clusters
- Share the answer as a map link
Put VOL AI in your product
Stream answers and citations with the SDK, or drop in the VolChat component and get the full experience in a few lines.
- SSE streaming
- Typed citations and clusters
- React and script embeds
import { VolClient } from "@vol/sdk";
const vol = new VolClient({ apiKey: process.env.VOL_API_KEY });
for await (const ev of vol.chatStream({ case: "papers", question })) {
if (ev.type === "delta") process.stdout.write(ev.text);
if (ev.type === "citations") console.table(ev.citations);
}Frequently asked questions
These are some of our most frequently asked questions.
Which model powers VOL AI?
VOL AI uses a retrieval-augmented pipeline built on RAGFlow with modern embedding, reranking and language models. Enterprise customers can bring their own models.
Can VOL AI hallucinate references?
Citations are generated from retrieved passages, not from the model's memory, so every reference points to a real document in your corpus.
Can I use VOL AI in my product?
Yes. Use the chat endpoint of the public API, the TypeScript SDK, the VolChat React component or the script embed.
Ask your first question today
- Product of the day1st
- Product of the week1st
- Product of the month1st