fix(azure): preserve Entra auth while polling OCR

[yujonglee-berri]

TLDR

Problem this solves:

• Regression from fix(azure_ai): preserve content, tables, and keyValuePairs in doc-intelligence /v1/ocr #32018 drops Entra polling authentication

How it solves it:

• Forward accepted request authentication to polling calls
• Cover subscription-key and Entra header forwarding

User Flow

Before: a developer uses Entra authentication for document OCR and the accepted analysis fails while LiteLLM polls for its result

1. They call litellm.ocr() with azure_ai/doc-intelligence/prebuilt-layout, an Entra token, and a document URL
2. Azure accepts the analysis request with 202 Accepted
3. LiteLLM polls without the bearer token and returns Unknown operation status: None

After: the same Entra-authenticated OCR call returns its completed document analysis

1. They call litellm.ocr() with azure_ai/doc-intelligence/prebuilt-layout, an Entra token, and a document URL
2. Azure accepts the analysis request with 202 Accepted
3. LiteLLM forwards the bearer token while polling and returns the completed OCR response

Relevant issues

Regression from #32018

Linear ticket

Pre-Submission checklist

• I have added meaningful tests
• The handful of test files covering my change pass locally
• My PR passes all required CI/CD checks (e.g., lint, schema.d.ts sync check, etc.)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review

Screenshots / Proof of Fix

Before (096984b)

1. Call litellm.ocr() against Azure Document Intelligence with an Azure CLI Entra bearer token and a public PDF
2. The analysis request returns 202 Accepted, but polling fails with Unknown operation status: None

After (40a01e5)

1. Run PYTHONPATH="$PWD" /Users/yujonglee/dev/litellm/.venv/bin/python -m pytest tests/test_litellm/llms/azure_ai/test_azure_document_intelligence_ocr_transformation.py -q
2. Output: 34 passed in 0.24s
3. Call litellm.ocr() against Azure Document Intelligence with the same Azure CLI Entra bearer token and public PDF
4. Output: POST Azure Document Intelligence OCR via Entra: model=doc-intelligence/prebuilt-layout, pages=2

Type

🐛 Bug Fix

Caveats (if any)

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

[codspeed]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing litellm_fix_azure_di_entra_polling (40a01e5) with litellm_internal_staging (096984b)}

[greptile-apps]

Greptile Summary

This PR preserves Azure Document Intelligence request authentication when polling accepted OCR operations

• Forwards either the subscription key or Entra authorization header to polling requests
• Adds regression coverage for both authentication methods
• Adds explicit parameter types requested by the previous review

Confidence Score: 5/5

The PR appears safe to merge, with the authentication regression covered and the previous typing requirement fully addressed

Polling now preserves either supported Azure authentication header without introducing new failures, and the manually resolved previous finding is fully fixed by explicit parameter annotations

Important Files Changed

Filename	Overview
litellm/llms/azure_ai/ocr/document_intelligence/transformation.py	Safely carries accepted-request authentication into SSRF-validated OCR polling requests
tests/test_litellm/llms/azure_ai/test_azure_document_intelligence_ocr_transformation.py	Covers subscription-key and Entra polling authentication with strongly typed parameters

_{Reviews (2): Last reviewed commit: "test(azure): type OCR polling auth cases" | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[yujonglee-berri]

@greptile-apps review