Skip to content

feat: add VideoPrism temporal video search - #117

Merged
tulayha merged 5 commits into
mainfrom
feat/videoprism-capability
Aug 13, 2026
Merged

feat: add VideoPrism temporal video search#117
tulayha merged 5 commits into
mainfrom
feat/videoprism-capability

Conversation

@tulayha

@tulayha tulayha commented Aug 13, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Add VideoPrism as a first-class temporal action-search capability shared by CLI, HTTP, MCP, and local worker paths.
  • Pin and prepare only the PyTorch VideoPrism checkpoint, expose temporal clip controls and search in Streamlit, and include the capability in desktop/runtime dependency metadata.
  • Flatten VideoPrism's pooled video and text outputs before Chroma storage and retrieval.
  • Existing indexes remain compatible, but media must be re-indexed with VideoPrism selected to use temporal search. The prepared model adds approximately 0.93 GiB to the local model cache.

Validation

  • uv run --no-sync pytest -q tests/test_videoprism.py — 4 focused capability tests passed.
  • uv run --no-sync pytest -q tests/test_frontend.py tests/test_frontend_app.py — 32 frontend tests passed.
  • uv run --no-sync ruff check src/vidxp/capabilities/videoprism tests/test_videoprism.py — capability lint passed.
  • Manual local boundary: prepared the pinned VideoPrism model, launched uv run --no-sync vidxp ui, and produced an active VideoPrism snapshot through Streamlit.

@tulayha
tulayha marked this pull request as ready for review August 13, 2026 11:12
@tulayha
tulayha merged commit 8922823 into main Aug 13, 2026
12 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant