Now in early access

Turn your docs into a
production-ready deep research API

Every report, proposal, and policy your company has ever written contains answers nobody can find. Dewey unlocks them — ask a question in plain English, get an answer in seconds with the exact passage it came from.

See how it works

No credit card required · Free tier available · Deploy in minutes

dewey api
POST /v1/collections/engineering-docs/ask
{
"question": "Which ADRs conflict with findings in the security audits?",
"depth": "research"
}
// Response
{
"answer": "ADR-012 and ADR-031 recommend client-side token storage, which the 2024 audit flags as a critical XSS risk...",
"citations": [{ document, section, chunk }]
}

Features

Everything the RAG stack needs.
Nothing it doesn't.

Real-time processing

Ingest anything

PDFs, DOCX, HTML, plain text. Every document becomes clean Markdown, structured sections, and embedded chunks — instantly queryable.

🔍Semantic + keyword

Retrieve precisely

Hybrid BM25 + vector search with RRF. Section-aware scanning finds the right document before paying for full retrieval.

💬With citations

Ask, get answers

One call to /ask returns grounded, cited answers. Configurable depth from quick lookups to deep multi-document research.

🔑RBAC built in

API-first from day one

Project-scoped API keys, org management, and role-based access. The infrastructure layer your whole team can rely on.

📡WebSocket + REST

Real-time events

WebSocket streams for every document lifecycle event — uploading, processing, sectioned, embedded, ready. Build reactive UIs instantly.

⚙️BYOK support

Bring your own model

Use Dewey's managed embeddings or plug in your own models with BYOK. Swap without re-indexing your entire corpus.

How it works

Ship AI search in three steps.

01

Create a project

Sign up, create an org, and provision a project in seconds. You get live and test API keys immediately — no infrastructure to configure.

POST /v1/projects
02

Upload documents

POST any file to a collection. Dewey converts it to Markdown, extracts a section hierarchy, chunks it, and embeds everything in real time.

POST /v1/collections/:id/documents
03

Query and ask

Query for hybrid search results with full citation lineage back to the source, or ask for a grounded answer synthesized at any research depth — from a quick summary to a deep cross-document analysis.

POST /v1/collections/:id/query
POST /v1/collections/:id/ask