5 commits

Author	SHA1	Message	Date
mubashir1osmani	999d23dfce	fix(hosted_vllm): forward Omni media URLs instead of downloading them ... OpenAI videos never fetches input URLs or invents download size caps. Drop MAX_VIDEO_MEDIA_* and the fetch-and-inline path. Keep MAX_IMAGE_URL_DOWNLOAD_SIZE_MB for chat image handling. Validate Omni reference URLs, then send them as form fields the way OpenAI videos sends input_reference as a file	2026-08-24 17:59:52 -04:00
mubashir1osmani	984b4b5968	fix(hosted_vllm): cap remote media downloads on video create ... Stream image, video, and audio URL bodies with Content-Length and byte caps so an authenticated caller cannot exhaust worker memory	2026-08-24 17:41:49 -04:00
mubashir1osmani	a5779151e0	fix(hosted_vllm): inline Omni media URLs through SSRF-safe fetch ... Fetch image, video, and audio reference HTTP URLs with safe_get and replace them with data URLs so Omni never requests user-controlled hosts	2026-08-24 17:31:34 -04:00
mubashir1osmani	01006b713a	refactor(hosted_vllm): register videos adapter via llms getter ... Keep Hosted VLLM video construction under litellm/llms and have ProviderConfigManager call the getter instead of importing the config class	2026-08-24 17:16:40 -04:00
mubashir1osmani	8f812f5636	feat(hosted_vllm): add vLLM-Omni videos API ... Route hosted_vllm video generation through /v1/videos as multipart form data so Omni extra fields such as width and extra_params reach the server instead of a JSON body Omni rejects	2026-08-24 17:10:44 -04:00