Real-time REST APIs for document parsing, extraction, and classification are now Generally Available
by Archika Dogra, Jane Zhang, Zachary Wang and Nihit Desai
At Databricks, our enterprise customers receive millions of documents every single day. These messy, unstructured documents hold critical context about operations, customers, and finances that power key downstream decisions across the business. Yet when a document arrives, putting that enterprise context to work is surprisingly tedious. For example:
- A contract lands in an inbox. A legal team member opens it, hunts for the parties, effective date, and renewal terms, then enters them into the system to track the agreement for years.
- An invoice arrives. An accounts payable specialist finds the vendor, PO number, and total, then enters them into the system so the invoice can go through approval and audit.
- An insurance claim shows up in a portal with a stack of attachments, and an adjuster reads all of it just to decide where the claim should be routed to.
These are distinct workflows, but they share the same bottleneck: before customers can use the valuable data inside the documents for business analytics or decision-making, someone has to read each document and pull out what matters. That manual step slows everything downstream, and the bottleneck and costs explode as the incoming document volume grows.
Process documents in real-time with synchronous REST APIs
Today, the Document Intelligence REST APIs for parsing, classification, and extraction are Generally Available. Our decisions REST API, great for document routing, is launched in beta. Developers can call these APIs from an application or agent when a document arrives, so the workflow can identify what came in, pull out the details it needs, and put them in front of the right person or agent for next steps.
These APIs transform your documents into structured insights in real-time:
- ai_parse_document reads the file, converts tables and figures into structured text, and calculates citation bounding boxes for downstream processing.
- ai_decide uses the document’s content to choose what happens next, such as routing the document to a workflow or flagging it for review.
- ai_extract pulls the fields you specify into a structured result, with optional citations and confidence scores, so humans or agents can check values against the source.
- ai_classify assigns labels to the document or extracted entities using your business taxonomy with optional rationales and confidence scores.
Example using the ai_extract REST API to extract fields from an invoice:
Integrate real-time document intelligence into your Apps and Agents
You can use the REST APIs to bring document understanding into the application where work already happens. For example, with Databricks Apps, you can build a contract intake flow where a lawyer uploads an agreement and opens a review form with the key terms already filled in:
- Read the contract.
ai_parse_document turns the file into structured content that the app can save for other uses.
- Open the right review flow.
ai_classify identifies whether the agreement is an NDA, MSA, or statement of work, so the app knows which form to show.
- Prefill the terms.
ai_extract pulls out the parties, effective date, renewal terms, and notice period. Citations let the lawyer check each value against the source.
- Move the work forward. The lawyer corrects or approves the fields, and the app writes the reviewed record and source reference to Delta.
Interactive, human-in-the-loop applications are just one way to leverage real-time document processing. You can also call these APIs from a custom agent built on Agent Bricks when it needs to understand or analyze incoming documents before making its next decision.
Get started today
Want to build a workflow like this? Open Genie Code in your Databricks workspace and start typing your prompt:
You can also build directly with the REST API reference and the guides for parsing, classification, and extraction:
Instagram Engineering
Shopify Engineering