hardikagarwal23/kuku_promo_studio

How hardikagarwal23/kuku_promo_studio works

Kuku PS1 turns screenplay text into ranked, event-boundary promo recommendations; it does not edit video. The CLI pipeline parses episodes, builds a show profile, generates and filters candidates, scores them with an LLM, optionally extends strong candidates, validates and ranks them, then exports results. A Streamlit review UI and studio operations provide a separate editorial workflow. The README documents the stages; source-index references and excerpts support the internal links shown here, while unsampled implementation details are not assumed.

Main entry points: run.py CLI Runner · llm_client.py LLM Client · app.py Review Dashboard · scoring.py Candidate Scorers

Components 20

Recommendation Pipeline

CLI Runner src/run.py
Screenplay Parser src/parser.py
Parsed Events src/schemas.py
Window Generator src/candidates.py
Recall Prefilter src/prefilter.py
Validation and Ranking src/validation_ranking.py
Result Exporter src/exporter.py

Narrative Analysis

Show Profiler src/show_profile.py
Candidate Scorers src/scoring.py
Start Extension src/extension.py
LLM Client src/llm_client.py

Run Operations

App Configuration src/config.py
Run Reliability src/reliability.py
Run Outputs

Editorial Review

Review Dashboard app.py
Dashboard Data dashboard_data.py
Studio Operations studio_ops.py

Other components

Screenplay Editor
LLM Provider
Human Editor

Connections 22

Screenplay Editor → CLI Runner: provides scripts README.md
CLI Runner → Screenplay Parser: parses episodes src/run.py
Screenplay Parser → Parsed Events: creates events src/parser.py
CLI Runner → Window Generator: generates windows src/run.py
CLI Runner → Recall Prefilter: filters candidates src/run.py
CLI Runner → Show Profiler: builds profile src/run.py
Show Profiler → LLM Client: requests profile src/show_profile.py
CLI Runner → Candidate Scorers: scores candidates src/run.py
Candidate Scorers → LLM Client: requests evaluations src/scoring.py
LLM Client → LLM Provider: sends prompts src/llm_client.py
CLI Runner → Start Extension: may extend src/run.py
Start Extension → Candidate Scorers: rescores extension src/extension.py
CLI Runner → Validation and Ranking: validates and ranks src/run.py
CLI Runner → Result Exporter: exports results README.md
CLI Runner → Run Outputs: writes run files README.md
CLI Runner → Run Reliability: records checkpoints src/run.py
LLM Client → App Configuration: loads settings src/llm_client.py
Review Dashboard → Dashboard Data: loads dashboard data app.py
Review Dashboard → Studio Operations: uses studio operations app.py
Human Editor → Review Dashboard: reviews promos README.md
Dashboard Data → Run Outputs: reads run data
Studio Operations → CLI Runner: starts pipeline studio_ops.py

Drawn by GitDiagram on November 12, 2026 from the repository's file tree, README and source files. Read as Markdown ·