VOXELL ANSWERS // FOR AGENTIC RAG DEVELOPERS
Stop building RAG glue. Voxell Answers handles the hard parts of document Q&A: chunking, Forge embeddings, Gemini enrichment, HNSW vector retrieval, BM25 hybrid search, and cited answer generation. Upload documents, ask questions, and update stale chunks without rebuilding your RAG stack.
50 ANSWERS / MONTH · 5,000 DOCUMENTS · NO CARD
Built on Forge embeddings · Hybrid dense + BM25 retrieval · Source-grounded answers · Editable chunks · Simple API updates
# upload docs, ask questions, get cited answers
> When can we terminate without penalty?
answer:
Either party may terminate within 30 days of renewal at no charge. [msa.md · chunk 2]
retrieved from:
6 documents · 13 chunks · hybrid dense + BM25
01 // WHEN TO USE IT
You need document answers without owning the retrieval infrastructure.
SKIP THE INFRASTRUCTURE
Answers handles upload, chunking, embedding, retrieval, citation, and answer generation. No separate vector DB. No chunking pipeline. No retrieval glue.
OPEN IT IN THE DASHBOARD →BETTER CONTEXT
Answers combines Forge embeddings, HNSW vector search, and BM25 hybrid retrieval so answers start from stronger context.
SEE RETRIEVAL FLOW →ENRICHMENT ON INGEST
Answers enriches content during ingestion to make tricky language easier to retrieve and answer correctly, including negations, exceptions, and policy-style clauses.
SEE HOW INGESTION WORKS →LIVING CORPORA
Edit chunks directly or update stale content with a simple curl call. Keep answers current without rebuilding the whole pipeline.
SEE UPDATE API →02 // MANAGED PATH
Most teams build RAG by wiring together five or six moving parts. Answers packages the whole path into one managed workflow.
03 // LIVE ENGINE
Sign in once, the free tier covers it, then drop in some text in the dashboard (release notes, a contract, a wiki page) and ask a question. You'll get an answer grounded in what you loaded, with the passage it came from. This is the real engine, not a canned demo.
VOXELL ANSWERS // LIVE DASHBOARD
Upload a document and query it against the live engine, free on the included allowance.
04 // PIPELINE
Three stages, one workflow: Mine your knowledge in, shape what stays, then ask it. The Forge stage shapes your corpus and is not to be confused with Voxell Forge, the embedding API it runs on.
01 / MINE
Upload text, files, or up to 256 documents in a single batch call. Voxell Answers splits them into clean, paragraph-aware chunks and embeds every one, no pre-chunking, no separate embedding step. As content comes in, Answers enriches it with Gemini so hard negations, exceptions, and policy-style clauses stay retrievable. Re-upload the same file and it's recognized, not duplicated.
02 / FORGE
Organize documents into corpora, rename and prune chunks, and score corpus quality with VQS to find the documents that hurt retrieval. Curate the knowledge that answers questions; drop the noise.
03 / ASK
Query in two modes: vectors returns the ranked passages; answer returns a Gemini-written response grounded in those passages, with citations back to the source chunk. One endpoint, one call.
05 // ENGINE CHOICE
Same corpus, same endpoints, only ?tier= changes. Start free and fast on Edge; switch to Precision when exact names, codes, and nuance matter.
EDGE
PRECISION · SOTA
06 // LIVING CORPORA
RAG systems get painful when the source material changes. Answers is designed for living corpora: update a stale chunk, patch content through the API, and keep asking questions without starting over.
curl -X PATCH https://api.voxell.ai/v1/answers/{corpus}/documents/{doc_id}/chunks/{index} \
-H "Authorization: Bearer $VOXELL_API_KEY" \
-H "Content-Type: application/json" \
-d '{"description":"Updated contract language..."}'# list the document's chunks first to get {doc_id} and {index}
07 // WHY ANSWERS
One managed path from raw documents to answers people can verify.
DROP-IN
Auto-chunking, embedding, retrieval, and the grounded answer are one product, not five services you integrate. Upload, then query. Citations come back with the answer, full provenance to the source chunk, no post-processing.
INGEST → QUERY, ZERO GLUE CODEQUALITY YOU CAN SEE
VQS (Voxell Quality Score) reads your stored chunks and flags the documents dragging retrieval down, near-duplicates, incoherent splits, junk attractors, with a per-document score and a 2D quality map. Stop guessing why a query missed.
PER-DOC SCORE · 2D QUALITY MAP →08 // RETRIEVAL BLUEPRINT
Answers doesn't just throw chunks at an LLM. It prepares documents for retrieval, embeds them with Forge, retrieves through an HNSW vector index, blends dense search with BM25, and sends grounded context to Gemini for the final answer, built on the same embedding models that put Voxell at the top of the MTEB leaderboard. Here's the engine underneath.
SPEC // 01
Precision embeds with Pro at 2560 dimensions, from Voxell's hard-negative-tuned embedding lineage that ranks at the top of MTEB. Retrieval quality is bounded by your embedding model; we lead with the part that's benchmarked.
SPEC // 02
Vectors live in a durable, managed store as half-precision values for compact, fast cosine search over an HNSW index. An iterative scan returns a full top-k even under tenant + corpus filters, so there's no silent under-retrieval.
SPEC // 03
A dense HNSW pass always runs; a BM25 lexical pass kicks in for queries carrying codes or identifiers. The two ranked lists are merged with RRF (reciprocal rank fusion), the recall of semantics with the precision of exact match.
09 // THE DROP-IN
Ingest a document, then query it in answer mode. The response is a grounded answer plus the chunks it cited, no chunking library, no vector store, no orchestration. The corpus is created on first upload.
TURBO FREE FOREVER · 50 ANSWERS / MONTH · NO CARD
# 1. load a document (auto-chunks + embeds)
curl https://api.voxell.ai/v1/answers/handbook/documents \
-H "Authorization: Bearer $VOXELL_KEY" \
-d '{"name":"policy.md","text":"..."}'
# 2. ask a question, get a grounded answer
curl https://api.voxell.ai/v1/answers/handbook/query \
-H "Authorization: Bearer $VOXELL_KEY" \
-d '{"query":"How many vacation days?","mode":"answer"}'
# => { "answer": "...", "chunks": [{ "doc_id": "...",
# "chunk_index": 3, "score": 0.87, "text": "..." }] }10 // PRICING
You pay per answer, $1.50 per 1,000, not a monthly rent on every vector sitting in storage. Turbo embeddings are free forever, and on the free plan your first 50 answers each month are on us. Past that you draw on prepaid credit, no subscription required.
FREE TO START
Turbo embeddings are free forever and never metered, and the corpus is the large half of the deal: 5,000 documents (60,000 vectors) of storage, with production use allowed. The answer allowance is the ordinary half, 50 a month. No card, no time limit. Your corpus stays live for 90 days after the last activity, and a question resets the clock.
NO CARD · 5,000 DOCUMENTSPAY PER ANSWER
There is no separate per-query charge. One answer, one price: $1.50 per 1,000, past the 50 the free plan includes. Retrieval and the query embed are inside it.
$0.0015 PER ANSWER · NO QUERY TOLLNO STORAGE RENT
There's no per-gigabyte, per-day charge on stored vectors. A large corpus you query occasionally doesn't quietly run up a bill the way per-storage pricing does.
EMBED ONCE · KEEP IT11 // CONCEPTS
The practical differences behind retrieval, grounding, and the two engine tiers.
For one document and one question, a chat window is fine. It falls apart the moment you have more documents than fit in a prompt, a support team with 50,000 articles, a product that answers from a live knowledge base, a contract set spanning years. You can't paste all of it into a chat.
Voxell Answers pre-processes your whole corpus into searchable meaning once, then retrieves only the few passages that matter when a question comes in. The LLM sees the right context instead of everything, so it's faster, cheaper, and it scales. And because retrieval quality is the ceiling on answer quality, Voxell Answers is built on embeddings that are benchmarked to be good.
Edge is fast, free, and great for everyday documents and broad questions, dense vector search on 1024-dimension embeddings. Precision is the accurate one: larger 2560-dimension embeddings, an HNSW index, and hybrid search that adds exact lexical matching for codes and identifiers, plus corpus quality scoring.
They share one API and one corpus model, you can start on Edge and move to Precision by changing a single tier parameter, no re-integration.
Pure semantic (dense) search is great at meaning but can blur exact tokens, it'll happily treat v2.1.4 and v2.4.1 as close. Pure keyword search nails exact strings but misses paraphrases.
Hybrid runs both and fuses the rankings. When your query contains an identifier, a SKU, an error code, a version, Precision adds a lexical pass so the exact match surfaces, while still using semantics for everything else. You get recall and precision without choosing.
In answer mode, Voxell Answers retrieves the most relevant chunks from your documents and asks Gemini to answer using only those, returning the source chunks alongside the answer. Every answer is traceable to the passage it came from. If nothing relevant is found, you get told that, not a confident guess.
12 // START
Start with the API, inspect the live product, or bring a difficult corpus for a private evaluation.
API
Use Answers as a managed document-to-answer layer. Upload content, query it, and update chunks through simple API calls. Wire it in with recipes or drive it from the forge-cli.
START FREE →LIVE
Create the free account, upload a document, ask a question, and see the cited answer. The 50 answers a month the free plan includes cover it.
START FREE →EVAL
Have a hard corpus, stale docs, contract clauses, policies, or support material? Send a sample and Voxell will show how Answers handles retrieval and citations.
REQUEST DOCUMENT EVAL →YOUR FIRST CITED ANSWER
Free to start: Turbo free forever, 5,000 documents (60,000 vectors) of corpus, 50 cited answers a month, no card required.