Video Knowledge Pipeline
How an approved teaching video becomes governed, retrievable evidence — verified against data/manifests/production_approval.json and data/indexes/strategies.json.
Approved source universe
Active source universe: OV-0001, OV-0002, OV-0003, OV-0004. Excluded from active runtime influence: OV-0005 (historical references remain audit-only).
| Source ID | Review status | Active runtime influence | Strategy references |
|---|---|---|---|
| OV-0001 | ai_cleaned_needs_validation | Yes | STRAT-0001, STRAT-0002, STRAT-0003 |
| OV-0002 | ai_cleaned_needs_validation | Yes | STRAT-0004, STRAT-0005, STRAT-0006, STRAT-0007, STRAT-0008 |
| OV-0003 | ai_cleaned_needs_validation | Yes | STRAT-0009, STRAT-0010, STRAT-0011, STRAT-0012, STRAT-0013, STRAT-0014 |
| OV-0004 | ai_cleaned_needs_validation | Yes | STRAT-0004, STRAT-0015, STRAT-0016, STRAT-0017, STRAT-0018, STRAT-0019 |
| OV-0005 | ai_cleaned_needs_validation | No — audit only | — |
Evidence: data/manifests/production_approval.json, data/indexes/strategies.json
Pipeline
CODE-DERIVED VISUALISATION
- 1Public video/source
- 2Transcript acquisition
- 3Raw preservation (data/raw/transcripts, append-only)
- 4Cleaning (src/cleaning/heuristic_extractor.py)
- 5Segmentation (src/chunking)
- 6Metadata attachment
- 7Concept extraction (src/extraction)
- 8Strategy mapping
- 9Semantic review
- 10Contradiction indexing
- 11Vector indexing (src/embeddings)
- 12Retrieval (src/retrieval/hybrid.py)
- 13Candidate evidence
This page never reproduces transcript text — only status, IDs, and strategy references derived from manifests.