feat(enrich): video frame descriptions + full transcript into enrich#75
Merged
Conversation
…rich Digested videos now contribute what they SHOW (key-frame descriptions) and the FULL spoken transcript to the summary/topics enrichment — not just a 12k-truncated transcript with frames ignored. - New shared `_video_frame_descriptions` helper (executors/api.py), threaded into both the api prompt (bounded by the new FRAME_DESC_CHAR_LIMIT, with a "frames omitted" marker) and the manual/claude-code worksheet (all frames). A mute screen-share video now contributes visual topic signal even with no speech. - The worksheet carries the FULL transcript (agent-judged, full context); the api engine keeps its TRANSCRIPT_CHAR_LIMIT bound. The two engines diverge on input size by design. Closes the "digested videos get a tweet-only summary" gap for the worksheet backfill. Gate green: 1107 tests, 95% cov. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Make digested videos get the same treatment as text: a content-based summary + topic assignment grounded in what the video says (ASR) and what it shows (frame descriptions).
The gap (verified in code)
A digested video carries a full transcript and per-frame descriptions, but
enrichonly saw a 12k-truncated transcript and never the frame descriptions (those live onx_video.frames[].description, a different field from theMediaPhotoDescribedphoto descriptions). So slide/screen-share videos with no speech contributed zero content signal → tweet-only summary/topics.Change
_video_frame_descriptions(item)helper (executors/api.py), threaded into both enrich engines:Video frames (slides/screens shown)block, bounded by the newFRAME_DESC_CHAR_LIMITwith a[… N further frames omitted …]marker.video_frame_descriptionsfield (all frames).has_speech=False.TRANSCRIPT_CHAR_LIMITbound. The two engines diverge on input size by design (documented).Why now
Unblocks the backfill of the 433 pending-enrich items (217 videos) via the worksheet engine ("our Claude") — see plan
zz-support-files/docs/implementation-plans/video-content-enrich.md.Tests
Gate
✅ 1107 tests · 95% cov · ruff/format/mypy/radon/bandit/vulture/interrogate/deptry all green.
🤖 Generated with Claude Code