Whisper fallback for YouTube + Knowledge Graph test isolation fix - #123
Merged
Matee-ur-Rehman merged 2 commits intoSep 12, 2026
Merged
Conversation
- Implemented Whisper-based transcription fallback for YouTube videos with disabled/unavailable captions (ingestion/youtube/whisper_fallback.py), wired into transcript.py's fetch_transcript(). Tested on both caption-disabled and normal videos. - Fixed a bug where knowledge_graph's test suite wrote synthetic 16-dimension vectors directly into the real, shared ChromaDB collection, permanently locking it to an incompatible dimension and breaking real ingestion (384-dim). Tests now use an isolated, in-memory Chroma client instead. Verified fix against real ingested documents post-recreation of the collection.
|
@usmanali434 is attempting to deploy a commit to the mrymfarid684-3634's projects Team on Vercel. A member of the Team first needs to authorize it. |
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Whisper Model — Implemented
Adds a fallback path in ingestion's YouTube transcript fetching: when a
video has no available captions, audio is downloaded (yt_dlp) and
transcribed locally with Whisper instead of failing/returning empty text.
Tested against both a caption-disabled video and a normal captioned video
to confirm the existing fast path still works unchanged.
Knowledge Graph — Bug Fixed
The graph service's test suite was inserting synthetic test vectors
directly into the real, shared ChromaDB collection instead of an isolated
test database. Since a Chroma collection's dimension is fixed by its
first insert, this locked the real collection to 16 dimensions —
incompatible with the real embedding model's 384-dim output — breaking
real ingestion for anyone using the shared store.
Fix: tests now use an ephemeral in-memory Chroma client, never touching
the real store. Verified the actual fix by deleting the corrupted
collection, re-ingesting real documents, and confirming /graph correctly
returns real nodes and edges (including correctly showing an unrelated
document with no false connections).
Known follow-up (not in this PR)
ingestion/main.py's JWT auth (from the earlier NV-2 fix) appears to be
missing from the current working copy — investigating separately whether
this was lost during a merge/reorg. Will follow up once confirmed.