Skip to content

docs(roadmap): video content vectorization brainstorm and roadmap#644

Open
Kneesal wants to merge 2 commits intomainfrom
docs/video-content-vectorization-roadmap
Open

docs(roadmap): video content vectorization brainstorm and roadmap#644
Kneesal wants to merge 2 commits intomainfrom
docs/video-content-vectorization-roadmap

Conversation

@Kneesal
Copy link
Copy Markdown
Member

@Kneesal Kneesal commented Apr 2, 2026

Summary

  • Adds brainstorm doc and 10 roadmap tickets (feat-037 through feat-046) for video content vectorization / scene-level recommendations
  • Phase 1 scope: English, Spanish, French — three languages to verify locale-aware dedup works (no cross-locale bleed)
  • Scene analysis approach: actual video segments via Gemini 2.5 Flash (not still frames), following Netflix/YouTube direction
  • LLM extracts structured signals: felt needs/themes (primary), bible verses, content, emotional tone, demographics
  • No human tags for similarity — CMS tags are unreliable, all signal from LLM extraction
  • Pure vector similarity for Phase 1; user-driven scoring deferred to Phase 2
  • Estimated cost: ~$600-$900 for Phase 1 (en/es/fr), ~$2K-$4K for full catalog
  • Schema: scene_embeddings table with themes[], bible_verses[], demographics[] columns
  • Frontend prototype (feat-046) renders recommendations on the existing Experience route with locale switching

Test plan

  • Review brainstorm doc for completeness and accuracy
  • Verify roadmap ticket dependencies are bidirectional and consistent
  • Confirm cost model assumptions are reasonable (refine after data audit)
  • Validate schema design supports locale-aware queries and demographic filtering

🤖 Generated with Claude Code

Kneesal and others added 2 commits April 2, 2026 02:16
… tickets

Scene-level video embeddings for cross-film recommendations, starting with
English-only prototype. Uses Gemini 2.5 Flash to describe scenes from
extracted frames + transcript, then embeds descriptions via existing
text-embedding-3-small pipeline into a separate pgvector scene_embeddings
table.

Adds:
- Requirements doc with phased rollout, storage schema, cost model, and
  technology research (Spotify RecSys 2025 validates this approach)
- Parent feature feat-037 plus 9 sub-tickets (feat-038 through feat-046)
  covering data audit, scene boundaries, descriptions, embeddings table,
  backfill worker, visual fusion, recommendation API, pipeline integration,
  and demo experience frontend
- Updates feat-009 blocks to include feat-037 dependency

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Phase 1: en/es/fr to verify locale dedup
- Use video segments via Gemini, not still frames
- Extract: felt needs/themes, bible verses, content,
  tone, demographics
- Locale-aware queries, no cross-locale bleed
- No human tags, pure vector similarity for Phase 1
- Cost model ~$600-$900 for video segment approach
- Schema: themes[], bible_verses[], demographics[]

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@railway-app
Copy link
Copy Markdown

railway-app bot commented Apr 2, 2026

🚅 Deployed to the forge-pr-644 environment in forge

4 services not affected by this PR
  • @forge/web
  • @forge/cms/db
  • @forge/cms
  • @forge/manager

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