Developer preview
Meet the extraction API.
Send a document. Get its contents back, with an explicit account of what needs another look.
This guide describes the current KenOCR API. You need a running KenOCR server and an issued API key. The examples use the local server at http://localhost:8710. This website does not process uploads or issue keys; public hosted access is not available yet.
Send your first document
Upload an image or PDF as multipart form data. Supply your key through X-API-Key or an Authorization: Bearer header.
curl http://localhost:8710/v1/extract \
-H "X-API-Key: $KEN_API_KEY" \
-F "file=@document.pdf" \
-F "mode=balanced"KEN_API_KEY is an environment variable holding a key issued by the operator of your KenOCR server. Replace document.pdf with your file.
Choose the right reading mode
| Mode | Use |
|---|---|
fast | Use a single reading path. A single read does not establish proof. |
balanced | Use the balanced extraction plan for the document. |
proof | Request independent cross-checks. A request for proof does not guarantee the result can be proven. |
Engine availability and document support depend on your server. Consult its interactive reference at /docs for the complete request schema and supported parameters.
Read the result carefully
| Output | What it tells you |
|---|---|
| Pages and rows | The extracted content. Preserve empty values, dashes and nulls as distinct facts. |
| Flags | Ambiguities, disagreements and other conditions that need attention. |
proven | Whether the required cross-checks actually established agreement. false does not by itself mean every value is wrong. |
proven_reason | Why proof was or was not established. |
Long-running requests can become jobs. Follow the returned job information and use /v1/jobs/{id} to retrieve status rather than assuming every response is a completed extraction.
Other useful endpoints
| Endpoint | Purpose |
|---|---|
GET /v1/health | Check server and pipeline availability. |
GET /v1/models | Inspect the available engine catalogue. |
POST /v1/compare | Compare engine readings. |
GET /openapi.json | Read the server’s complete machine-readable API schema. |
GET /docs | Open the interactive API reference on your server. |
Authentication and scopes vary by endpoint. Your server’s API reference is the source of truth.
What’s next
KenOCR is being developed towards fast, affordable specialist APIs for invoices, handwriting and image-based PDFs. Those dedicated APIs and public pricing are future work; the current preview uses the extraction endpoint described here.