From 2df6dced0a6d3f7701a05aa0d8ad9f6811e323db Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Wed, 15 Jul 2026 19:44:29 +0000
Subject: [PATCH] fix(videos): re-stamp status response id with resolved
 model_id so downstream status/content resolve per-model key

---
 litellm/proxy/video_endpoints/endpoints.py    |  7 ++++++-
 .../proxy/video_endpoints/test_endpoints.py   | 21 +++++++++++++++++++
 2 files changed, 27 insertions(+), 1 deletion(-)

diff --git a/litellm/proxy/video_endpoints/endpoints.py b/litellm/proxy/video_endpoints/endpoints.py
index a13e80809c1..18a6acfe4e1 100644
--- a/litellm/proxy/video_endpoints/endpoints.py
+++ b/litellm/proxy/video_endpoints/endpoints.py
@@ -279,7 +279,7 @@ async def video_status(
     # Process request using ProxyBaseLLMRequestProcessing
     processor = ProxyBaseLLMRequestProcessing(data=data)
     try:
-        return await processor.base_process_llm_request(
+        response = await processor.base_process_llm_request(
             request=request,
             fastapi_response=fastapi_response,
             user_api_key_dict=user_api_key_dict,
@@ -297,6 +297,11 @@ async def video_status(
             user_api_base=user_api_base,
             version=version,
         )
+        return encode_video_id_in_response(
+            response=cast(object, response),
+            fallback_provider=custom_llm_provider,
+            fallback_model_id=model_id_from_decoded,
+        )
     except Exception as e:
         raise await processor._handle_llm_api_exception(
             e=e,
diff --git a/tests/test_litellm/proxy/video_endpoints/test_endpoints.py b/tests/test_litellm/proxy/video_endpoints/test_endpoints.py
index 12649579652..e64cc603259 100644
--- a/tests/test_litellm/proxy/video_endpoints/test_endpoints.py
+++ b/tests/test_litellm/proxy/video_endpoints/test_endpoints.py
@@ -367,6 +367,27 @@ async def test_status__header_provider_beats_decoded_id(harness):
     assert data["model"] == "azure-sora"
 
 
+@pytest.mark.asyncio
+async def test_status__reencodes_response_id_preserving_model_id(harness):
+    """Regression for #33423: the status transform re-encodes the video id with
+    model_id=None, so the id the client gets back decodes to an empty model_id.
+    Re-using that id on a later status/content call then cannot resolve the
+    deployment (or its per-model api_key) and dies with invalid_api_key. The
+    endpoint must re-stamp the response id with the model_id it decoded from the
+    incoming id so the returned id round-trips."""
+    stripped_id = encode_video_id_with_provider("video_orig123", "azure", "")
+    assert decode_video_id_with_provider(stripped_id)["model_id"] == ""
+    harness.base_process.return_value = _video_response(stripped_id, hidden_params={})
+
+    resp = await call_status(harness, AZURE_VIDEO_ID)
+
+    decoded = decode_video_id_with_provider(resp.id)
+    assert decoded["model_id"] == VIDEO_MODEL_ID
+    assert decoded["custom_llm_provider"] == "azure"
+    assert decoded["video_id"] == "video_orig123"
+    assert harness.resolve_model.side_effect(decoded["model_id"]) == "azure-sora"
+
+
 # =========================================================================== #
 #   GET /v1/videos/{video_id}/content  -  video_content                        #
 # =========================================================================== #