Document Intelligence API
Add documents to a collection, then search them, ask questions answered with cited sources, and read the fields extracted from every document.
Document Intelligence reads every document you add to a collection — PDFs, Word files, text, Markdown, audio, web pages — splits it into searchable sections, and extracts the fields you defined for that collection (an invoice's vendor, total and due date; a résumé's skills and experience). The API lets your code feed documents in and get answers, passages and structured data out.
What you can build
- Ingestion from your own systems. Push documents from a form, a mailbox, a storage bucket or a nightly job as they arrive.
- Questions with sources. Ask in plain language and show the answer with the documents it came from.
- Structured data out. Read extracted fields per document or as one table across the collection, and load them into a spreadsheet or database.
- Retrieval for your own model. Use search to fetch the most relevant passages and send them to the model of your choice.
How it fits together
| Concept | What it is |
|---|---|
| Collection | A set of documents with one field schema. Its id (a UUID) is the knowledge_id in every path. You create collections in the app. |
| Document | One file, text or web page in a collection, identified by an object_id. It is processed in the background after you add it. |
| Extracted fields | The values read out of each document according to the collection's fields — for example vendor_name, total_amount, due_date. |
| Section | A passage of a document, as indexed for search (the API calls these chunks). |
| API key | A secret (sk-…) sent in the x-api-key header, usually created for one collection. |
Base URL
https://futuresmart.ai/document-intelligence/apiEvery path in these docs is relative to it, for example POST /v1/{knowledge_id}/chat. The application's own address, https://app.futuresmart.ai/api/knowledge, serves the same routes.
A self-hosted installation of the open-source project serves the same routes from its own root.
Start here
- Quickstart — create a collection, get a key, add a document and ask about it.
- API keys and permissions — what a key can do and how to keep it safe.
- Add documents — every way in, size limits, and processing status.
- API reference — all endpoints, generated from the deployed OpenAPI contract.