test(e2e): cover fal Seedance video create, poll and download

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

tests/e2e/coverage_registry/llm_nonconversational.yaml

@@ -80,6 +80,7 @@
 - {id: llm.images_generations.vertex.basic.nonstream.works, module: llm, tier: P1, subject_endpoint: images_generations, route: vertex, capability: basic, streaming: nonstream, assertions: [works], source: "vertex_ai/image_generation/image_generation_handler.py", rationale: "Vertex Imagen"}
 - {id: llm.images_generations.bedrock.basic.nonstream.works, module: llm, tier: P1, subject_endpoint: images_generations, route: bedrock_converse, capability: basic, streaming: nonstream, assertions: [works], source: "bedrock/image_generation/image_handler.py", rationale: "Bedrock Titan Image"}
 - {id: llm.images_generations.black_forest_labs.basic.nonstream.works, module: llm, tier: P1, subject_endpoint: images_generations, route: openai, capability: basic, streaming: nonstream, assertions: [works], source: "black_forest_labs/image_generation/handler.py", rationale: "BFL Flux via OpenAI-compat"}
+- {id: llm.videos.fal_ai.basic.nonstream.works, module: llm, tier: P1, subject_endpoint: videos, route: fal_ai, capability: basic, streaming: nonstream, assertions: [works], source: "test_video_generation_e2e.py", rationale: "fal queue video create, poll, content download"}
 - {id: llm.audio_speech.openai.basic.nonstream.works, module: llm, tier: P1, subject_endpoint: audio_speech, route: openai, capability: basic, streaming: nonstream, assertions: [works], source: "test_audio_speech_e2e.py:22", rationale: "OpenAI TTS binary audio"}
 - {id: llm.audio_speech.openai.basic.stream.works, module: llm, tier: P1, subject_endpoint: audio_speech, route: openai, capability: basic, streaming: stream, assertions: [works], source: "proxy_server.py:9043", rationale: "TTS streaming chunk generator"}
 - {id: llm.audio_speech.openai.input_validation.nonstream.works, module: llm, tier: P1, subject_endpoint: audio_speech, route: openai, capability: input_validation, streaming: nonstream, assertions: [works], source: "vendor strategy §9.6 / LIT-4778", rationale: "TTS missing input/model, invalid voice, empty input rejected"}

tests/e2e/coverage_registry/schema.py

@@ -44,6 +44,7 @@ LlmEndpoint = Literal[
     "vector_stores",
     "ocr",
     "bedrock_native",
+    "videos",
 ]
 
 LlmRoute = Literal[
@@ -53,6 +54,7 @@ LlmRoute = Literal[
     "bedrock_converse",
     "bedrock_invoke",
     "cohere",
+    "fal_ai",
     "gemini",
     "hosted_vllm",
     "openai",

tests/e2e/llm_translation/LLM_TRANSLATION_COVERAGE_MATRIX.md

@@ -48,6 +48,7 @@ most likely to silently break and the one a mock can't prove works.
 |----------|---------------|-----------|------------|-------------|--------|
 | Chat | live (spend suite) | live (spend suite) | gap | live | partial |
 | Embeddings | live (spend suite) | n/a | n/a | live | covered |
+| Video | live (fal.ai Seedance) | n/a | n/a | - | partial |
 | Responses / image / audio / rerank / realtime | - | - | - | - | gap |
 
 ## This suite's files
@@ -61,6 +62,7 @@ most likely to silently break and the one a mock can't prove works.
 | `test_anthropic_passthrough_streaming_logs_cost` | anthropic native, stream, cost |
 | `test_anthropic_passthrough_tool_call_logs_cost` | anthropic native, tool call, cost |
 | `test_vertex_passthrough_via_managed_model_logs_cost` | vertex_ai native, non-stream, cost |
+| `test_fal_seedance_video_completes_and_downloads` | fal.ai Seedance video create, poll, and content download |
 
 Vertex keeps the credential on the proxy like gemini/anthropic, but the deployment is
 added at runtime instead of declared in the gateway config: the test POSTs `/model/new`

tests/e2e/llm_translation/endpoints_client.py

@@ -13,7 +13,7 @@ from dataclasses import dataclass
 from typing import Literal
 
 from e2e_config import SLOW_PROVIDER_TIMEOUT_SECONDS
-from e2e_http import BinaryStream, Result, StreamingResponse
+from e2e_http import BinaryStream, NoBody, Result, StreamingResponse
 from models import CacheControl, ChatMessage, LiteLLMParamsBody, RichMessage, TextBlock
 from proxy_client import ProxyClient
 from pydantic import BaseModel
@@ -26,6 +26,8 @@ __all__ = [
     "TextBlock",
     "TranscriptionForm",
     "TranscriptionResult",
+    "VideoObject",
+    "VideoRequest",
 ]
 
 
@@ -127,6 +129,13 @@ class ImageRequest(BaseModel):
     size: str = "1024x1024"
 
 
+class VideoRequest(BaseModel):
+    model: str
+    prompt: str
+    seconds: str = "4"
+    size: str = "1280x720"
+
+
 class ImageEditForm(BaseModel):
     model: str
     prompt: str
@@ -267,6 +276,12 @@ class ImagesResult(BaseModel):
     data: list[ImageItem] = []
 
 
+class VideoObject(BaseModel):
+    id: str
+    status: str
+    model: str | None = None
+
+
 class TranscriptionResult(BaseModel):
     text: str = ""
 
@@ -440,6 +455,25 @@ class EndpointsClient:
             "/v1/images/generations", key, ImageRequest(model=model, prompt=prompt)
         )
 
+    def videos(self, key: str, model: str, prompt: str) -> StreamingResponse:
+        return self._send(
+            "/v1/videos", key, VideoRequest(model=model, prompt=prompt)
+        )
+
+    def video_status(self, key: str, video_id: str) -> Result[VideoObject]:
+        return self.proxy.transport.get(
+            f"/v1/videos/{video_id}",
+            headers=self.proxy.transport.bearer(key),
+            params=NoBody(),
+            response_type=VideoObject,
+        )
+
+    def video_content(self, key: str, video_id: str) -> StreamingResponse:
+        return self.proxy.transport.download(
+            f"/v1/videos/{video_id}/content",
+            headers=self.proxy.transport.bearer(key),
+        )
+
     def image_edit(
         self, key: str, model: str, prompt: str, image: bytes, *, filename: str = "image.png"
     ) -> Result[ImagesResult]:

tests/e2e/llm_translation/test_video_generation_e2e.py

@@ -0,0 +1,69 @@
+"""Live e2e: POST /v1/videos creates a video and serves its content.
+
+Registers a fal.ai Seedance deployment at runtime, polls the queued video until it
+completes, and asserts the generated content is returned as binary data.
+"""
+
+from __future__ import annotations
+
+import time
+from typing import Final
+
+import pytest
+from e2e_config import unique_marker
+from e2e_http import require_successful_call, unwrap
+from endpoints_client import EndpointsClient, VideoObject
+from lifecycle import ResourceManager
+from models import LiteLLMParamsBody
+
+pytestmark = pytest.mark.e2e
+
+_POLL_INTERVAL_SECONDS: Final[float] = 5.0
+_POLL_TIMEOUT_SECONDS: Final[float] = 600.0
+
+
+def _wait_for_completion(
+    endpoints_client: EndpointsClient, key: str, created: VideoObject
+) -> VideoObject:
+    deadline = time.monotonic() + _POLL_TIMEOUT_SECONDS
+    while time.monotonic() < deadline:
+        status = unwrap(endpoints_client.video_status(key, created.id))
+        assert status.id == created.id
+        if status.status == "completed":
+            return status
+        if status.status == "failed":
+            pytest.fail(f"fal.ai video generation failed: {status}")
+        time.sleep(_POLL_INTERVAL_SECONDS)
+    pytest.fail(f"fal.ai video {created.id!r} did not complete within {_POLL_TIMEOUT_SECONDS}s")
+
+
+class TestVideoGeneration:
+    @pytest.mark.covers("llm.videos.fal_ai.basic.nonstream.works")
+    def test_fal_seedance_video_completes_and_downloads(
+        self, endpoints_client: EndpointsClient, resources: ResourceManager
+    ) -> None:
+        model = f"e2e-fal-video-{unique_marker()}"
+        model_id = endpoints_client.create_model(
+            model,
+            LiteLLMParamsBody(
+                model="fal_ai/bytedance/seedance-2.5/text-to-video",
+                api_key="os.environ/FAL_AI_API_KEY",
+            ),
+        )
+        resources.defer(lambda: endpoints_client.delete_model(model_id))
+        key = resources.key()
+
+        result = endpoints_client.videos(
+            key, model, "a red fox running through snow at dawn"
+        )
+        require_successful_call(result)
+        created = VideoObject.model_validate_json(result.body)
+        assert created.id
+        assert created.model
+
+        _wait_for_completion(endpoints_client, key, created)
+
+        content = endpoints_client.video_content(key, created.id)
+        require_successful_call(content)
+        assert len(content.body) > 0
+        assert not (content.content_type or "").startswith("application/json")