Skip to content

Latest commit

 

History

History
33 lines (24 loc) · 1.96 KB

File metadata and controls

33 lines (24 loc) · 1.96 KB

ScriptProof project rules

ScriptProof is a new submission for the Agentic Cinema hackathon, Parallel track. The submission deadline is 2026-09-10 05:00 Asia/Singapore.

Read first

  1. NEXT.md for current status and the next safe action.
  2. DEV_LOG.md for implementation decisions and evidence.
  3. docs/COMPETITION_RULES.md before changing the stack or submission scope.

Competition gates

  • This must remain a new, independent project. Do not turn Agentic Brief into this product or copy its repository history.
  • AI and agent tooling may use only Google Cloud AI plus Parallel's allowed capabilities. Do not add OpenAI, Anthropic, AWS AI, or another agent framework.
  • The production path must call Parallel Search at runtime. A README mention, mock, or cached result does not satisfy the track requirement.
  • The product must use Gemini and Google ADK / Agent Builder.
  • Final delivery requires a hosted app, public open-source repository, English project description, and a public demo video no longer than three minutes.

Product contract

  • Target users are screenwriters and pre-production teams.
  • The core flow is: read screenplay → identify research claims → call Parallel → preserve citations → check continuity → produce revision actions.
  • Supported and contradicted findings require at least one working HTTPS source.
  • Uncertain evidence stays uncertain. Never fabricate, repair, or silently replace citations.
  • Production has no fake-data fallback. Test doubles are allowed only in tests and visual QA.

Engineering constraints

  • Python 3.12, Google ADK, Gemini 3.5 Flash, Flask, Parallel's official parallel-google-adk package.
  • Vertex AI location is global.
  • Secrets stay in .env locally and Secret Manager in production. Never commit them.
  • Keep the paid analysis endpoint bounded and optionally protected by SCRIPT_PROOF_ACCESS_CODE.
  • All changes must pass Ruff, pytest, 80%+ coverage, interface QA, and an independent pre-push review.