Extract Document

POST /workspaces/{slug}/schemas/{schemaKey}/extract

Two modes of document resolution: 1. Existing document: Provide documentId to extract from a previously uploaded document. 2. Inline upload: Provide name + content (JSON) or name + file (multipart) to create a new document and extract in one request. Two modes of schema resolution: A. Use schemaKey from URL path to resolve the latest (or specific) version. B. Provide schema in body to auto-version (creates family/version if needed). Returns 202 with a runId. Poll GET /runs/{runId} for results.

Path Parameters

slug string required

schemaKey string required

Query Parameters

env "production" | "staging" | "development"

Environment slug. Defaults to production.

Body Parameters

documentId string

ID of an existing document to extract from

versionNumber integer

schema object

schemaName string

Display name for the schema family

description string

Description for the schema family

name string

Document name (for inline upload without documentId)

content string

Document content (for inline upload without documentId)

model string

Optional provider/model override for this run. Format is provider/model — for example anthropic/claude-opus-4-7. Takes precedence over the environment assignment. Returns 412 no_key if the named provider has no user key and is not Mistral.

curl -X POST \
-H "x-api-key: dsk_your_api_key" \
-H "Content-Type: application/json" \
-d '{
"documentId": "doc_abc123",
"versionNumber": 1,
"schema": {},
"schemaName": "example-schemaName",
"description": "Invoice processing workspace",
"name": "My Workspace",
"content": "Invoice #1234...",
"model": "example-model"
}' \
https://api.tracore.io/workspaces/my-workspace/schemas/invoice/extract?env=production

import { TracoreClient } from '@tracore/sdk';

const client = new TracoreClient({ apiKey: 'dsk_your_api_key' });

// Extract from existing document with auto-polling
const run = await client.extract('my-workspace', 'invoice', {
documentId: 'doc_abc123',
}, { poll: true });

if (run.status === 'completed') {
console.log(run.extractedData);
}

202 Response

{
  "runId": "run_abc123",
  "status": "pending"
}