Bring Your Own OCR

DocuDevs can extract text using your organization's own Azure Document Intelligence (Azure OCR) account instead of the DocuDevs platform account.

Who can manage OCR providers

Creating, editing, deleting, or key-binding OCR providers is not gated to organization admins today: any member of the organization holding an API key can perform these actions. This differs from embedding providers and model credentials, which require the ADMIN role for the same operations.

OCR providers do not support shared model credentials: each provider stores its own apiKey directly. This is different from LLM and embedding providers, which can reference a reusable model credential.

There is no dedicated verify/probe endpoint for OCR providers today, unlike POST /llm/providers/{id}/verify for LLM providers. Confirm a provider works by running a document-processing job against it.

Configure a provider

1. Open Settings → OCR Providers and select New.
2. Enter a name, the Azure Document Intelligence endpoint, the API key, and the model ID (for example prebuilt-document).
3. Optionally set provider-specific features.
4. Save the provider.
5. Bind the provider to one or more logical OCR keys.

Only explicitly bound keys use the provider. Unbound keys continue to use the DocuDevs platform OCR account.

Create a provider

cURL

API_URL="https://api.docudevs.ai"
API_KEY="<your-api-key>"

curl -X POST "$API_URL/ocr/providers" \
  -H "Authorization: $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Our Doc Intelligence",
    "endpoint": "https://myacct.cognitiveservices.azure.com",
    "apiKey": "<azure-doc-intelligence-key>",
    "modelId": "prebuilt-document"
  }'

Python SDK

The hand-written DocuDevsClient.create_ocr_provider/update_ocr_provider convenience methods send outdated field names (model instead of modelId) and will be rejected by the API. Until they are fixed, call the generated, typed functions directly:

from docudevs.client import AuthenticatedClient
from docudevs.api.azure_ocr_providers.create_ocr_provider import asyncio_detailed as create_ocr_provider
from docudevs.models.create_ocr_provider_request import CreateOcrProviderRequest

client = AuthenticatedClient(base_url="https://api.docudevs.ai", token="<your-api-key>")

body = CreateOcrProviderRequest(
    name="Our Doc Intelligence",
    endpoint="https://myacct.cognitiveservices.azure.com",
    api_key="<azure-doc-intelligence-key>",
    model_id="prebuilt-document",
)
response = await create_ocr_provider(client=client, body=body)
provider = response.parsed

Update or delete a provider

cURL

curl -X PATCH "$API_URL/ocr/providers/7" \
  -H "Authorization: $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"modelId": "prebuilt-invoice"}'

curl -X DELETE "$API_URL/ocr/providers/7" -H "Authorization: $API_KEY"

Python SDK

from docudevs.api.azure_ocr_providers.update_ocr_provider import asyncio_detailed as update_ocr_provider
from docudevs.api.azure_ocr_providers.delete_ocr_provider import asyncio_detailed as delete_ocr_provider
from docudevs.models.update_ocr_provider_request import UpdateOcrProviderRequest

body = UpdateOcrProviderRequest(model_id="prebuilt-invoice")
await update_ocr_provider(id=7, client=client, body=body)

await delete_ocr_provider(id=7, client=client)

Key bindings

Each logical OCR key is bound to a provider independently. Listing and binding keys use the correctly-mapped list_ocr_keys/update_ocr_key_binding convenience methods on DocuDevsClient.

cURL

curl "$API_URL/ocr/keys" -H "Authorization: $API_KEY"

curl -X PUT "$API_URL/ocr/keys/DEFAULT" \
  -H "Authorization: $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"providerId": 7}'

Python SDK

from docudevs import DocuDevsClient

client = DocuDevsClient(api_url="https://api.docudevs.ai", token="<your-api-key>")

keys = (await client.list_ocr_keys()).json()

await client.update_ocr_key_binding("DEFAULT", 7)
# Pass provider_id=None to clear a binding

list_ocr_keys/update_ocr_key_binding call the API with a plain httpx.AsyncClient under the hood and return a regular httpx.Response (.json(), .status_code), unlike the generated functions used above.

LLM providers

Chat/extraction models are configured separately under Settings → LLM Providers — see Bring your own LLM. OCR and LLM bindings are independent of each other.