feat: generic OpenAI-compatible video custom provider

[oyi77]

Summary

Adds a generic OpenAI-compatible video generation path so users can add custom video providers (base URL + API key) without per-provider code changes.

Changes

Core Implementation

• open-sse/handlers/videoGeneration/openai.ts (new): Generic handler with:

  • resolveVideoEndpoint() - resolves endpoint URL from credentials (prefers providerSpecificData.baseUrl)
  • fetchVideoEndpoint() - fetches with timeout, error handling, and logging
  • handleOpenAIVideoGeneration() - main handler accepting duration parameter

• open-sse/handlers/videoGeneration.ts:

  • Added resolveVideoBaseUrl() helper (mirrors resolveImageBaseUrl)
  • Added import for handleOpenAIVideoGeneration
  • Added dispatch for openai-video format before vertex-veo
  • Updated synthetic config to use format: "openai-video" for custom providers
  • Fixed fallback to only use OpenAI handler for resolvedProvider (custom models), returning Unsupported video format for unknown built-in formats
  • Restored adobe-firefly-video dispatch branch

• src/app/api/v1/videos/generations/route.ts: Already had correct custom model logic:

  • Imports getAllCustomModels
  • Scans custom models for supportedEndpoints.includes("videos")
  • Parses providerId/modelId from custom model registry
  • Resolves credentials via getProviderCredentialsWithQuotaPreflight(provider, null, null, requestedModel)
  • Passes resolvedProvider to handler

• src/shared/validation/schemas/provider.ts: Added "videos" to supportedEndpoints enum

Tests

• tests/unit/video-generation-handler.test.ts: Added handler-level test for custom provider
• tests/unit/video-custom-provider-route.test.ts (new): Route-level tests covering:
  • Custom provider with videos endpoint (happy path)
  • Custom provider without videos endpoint (400)
  • Unknown custom provider (400)

Design Decisions (Locked)

• Video only; model accepts duration
• Custom models tagged "videos" in supportedEndpoints
• Handler signature gains resolvedProvider (images parity)
• New format "openai-video" dispatched FIRST (before vertex-veo)
• Route builds fallback providerConfig {id, baseUrl, authType:"apikey", authHeader:"bearer", format:"openai-video"} when !providerConfig && resolvedProvider
• Credentials via getProviderCredentialsWithQuotaPreflight(provider, null, null, requestedModel)
• Call-log shape mirrors existing video handlers
• Error text sanitized via sanitizeErrorMessage
• fetchWithTimeout from @/shared/utils/fetchTimeout (no .ts)

Verification

• npm run typecheck:core passes
• 17 video tests pass (3 new route tests + 1 new handler test + 13 existing)
• No regressions in image generation tests

Verification

• npm run typecheck:core — PASS (zero output, exit 0)
• tests/unit/video-custom-provider-route.test.ts — 3/3 pass
• tests/unit/video-generation-handler.test.ts — 14/14 pass
• tests/unit/image-generation-handler.test.ts — 43/43 pass (no regression)

Coverage (c8, isolated output dirs)

Suite	File	stmts	funcs	lines
route	open-sse/handlers/videoGeneration/openai.ts	85.89%	100%	85.89%
handler	open-sse/handlers/videoGeneration/openai.ts	84.61%	100%	—
handler	open-sse/handlers/videoGeneration.ts	62.53%	—	—
handler	open-sse/handlers/imageGeneration.ts	68.32%	—	—

Scoping note (official Sora SDK)

The openai-video handler is a gateway-house convention for OpenAI-compatible video
providers — POST /v1/videos/generations → synchronous MP4 response, credentials from the
custom-provider connection (baseUrl + API key). This is not the official OpenAI SDK
Sora job API (POST /v1/videos → GET /videos/{id} polling). Sora-style async job providers
are planned in the follow-up mapping layer (generationConfig presets + job/poll branch).

[diegosouzapw]

Thanks for the PR. Please address the mandatory items (tests and/or merge blockers) in this branch, then rerun checks before /merge-prs.

[diegosouzapw]

Thanks for the PR. Please address the mandatory items (tests and/or merge blockers) in this branch, then rerun checks before /merge-prs.

[diegosouzapw]

Thanks for the PR. Please address the mandatory items (tests and/or merge blockers) in this branch, then rerun checks before /merge-prs.

[diegosouzapw]

Obrigado pelo PR. Mantive a revisão de fix-in-place e não foi possível concluir o ajuste completo aqui:

• Para os PRs em fork: não consigo aplicar push de correção diretamente na sua branch.
  Por favor, faça um rebase/sync com release/v3.8.50, resolva conflitos se houver, e rode os checks dessa branch.
  Se preferir, posso aplicar a correção na próxima rodada assim que você mandar o branch atualizado ou confirmar que o PR está limpo pra esse merge.

[mergify]

⚠️ The sha of the head commit of this PR conflicts with #8875. Mergify cannot evaluate rules on this PR. Once #8875 is merged or closed, Mergify will resume processing this PR. ⚠️

[oyi77]

Closing as superseded — the feature content is already merged upstream:

• Cherry-picked into release/v3.8.50 as 580162548 (cherry-pick(pr-9818): feat: generic OpenAI-compatible video custom provider #9844), with follow-up maint fixes maint: follow-up cherry-pick fix-in-place #9629 (conflict-resolved fallback) #9885 / maint: follow-up cherry-pick fix-in-place #9704 (conflict-resolved fallback) #9889 / maint: follow-up cherry-pick fix-in-place #9711 (conflict-resolved fallback) #9891.
• 9 of 11 PR files are byte-identical in base; the remaining two differ only by upstream's own refactor (Runway helpers extracted to videoGeneration/runwayHelpers.ts, ok === false → !ok in job.ts) — semantically identical.
• Per the maintainer's request, the fork head feat/generic-openai-video-provider has been synced to release/v3.8.50 tip 382449d59 (was 109 commits behind).