How colingpt9/grand-theft-auto-6-clips-kitty-plugin works
This is a Clips Kitty plugin that finds and ranks GTA highlight moments from a stream transcript and host-provided media signals, then returns clip ranges and optional edit suggestions to the host. The workflow runs through the SDK entry point and adapter, which gathers evidence and invokes detection before submitting results. Separate branches cover dialogue review with an optional local model and evaluation tooling. The map follows the README and sampled excerpts; some internal call details and unsampled implementations are not shown, so uncertain links are omitted.
Main entry points: adapter.py Job adapter · detector.py Highlight detector · run.py Fixture evaluator · dialogue.py Dialogue discovery
Components 16
Host integration
- Clips Kitty
- SDK entry point
src/main.py - Job adapter
src/gta_highlights/adapter.py
Moment detection
- Evidence collection
src/gta_highlights/observations.py - Evidence and scoring
src/gta_highlights/scoring.py - Event taxonomy
src/gta_highlights/taxonomy.py - Highlight detector
src/gta_highlights/detector.py - Candidate selection
src/gta_highlights/selection.py
Dialogue review
- Dialogue discovery
src/gta_highlights/dialogue.py - Review budget and cache
src/gta_highlights/review.py
Evaluation
- Fixture evaluator
evaluation/run.py - Energy baseline
evaluation/baseline.py - Temporal metrics
evaluation/metrics.py
Other components
- Streamer
- Stream VOD
- Local Ollama model
Connections 17
- Streamer → Stream VOD: provides
README.md - Stream VOD → Clips Kitty: adds VOD
README.md - Clips Kitty → SDK entry point: invokes plugin
src/main.py - SDK entry point → Job adapter: runs
src/main.py - Job adapter → Evidence collection: collects evidence
src/gta_highlights/adapter.py - Evidence collection → Evidence and scoring: creates records
src/gta_highlights/observations.py - Job adapter → Highlight detector: detects highlights
src/gta_highlights/adapter.py - Highlight detector → Event taxonomy: classifies cues
src/gta_highlights/detector.py - Highlight detector → Evidence and scoring: scores support
src/gta_highlights/detector.py - Job adapter → Candidate selection: uses selection
src/gta_highlights/adapter.py - Job adapter → Dialogue discovery: reviews dialogue
src/gta_highlights/adapter.py - Dialogue discovery → Review budget and cache: uses review limits
src/gta_highlights/dialogue.py - Dialogue discovery → Local Ollama model: requests review
- Job adapter → Clips Kitty: submits clip ranges
README.md - Fixture evaluator → Highlight detector: evaluates
evaluation/run.py - Fixture evaluator → Energy baseline: compares baseline
evaluation/run.py - Fixture evaluator → Temporal metrics: measures results
evaluation/run.py