CBCreativeBenchEnterprise creative-writing evaluation

Submission intake

Test the receipt path with a real alpha submission.

Import a manuscript file or paste text directly. The current API runs a deterministic dry-run judge panel first, so we can test storage, receipts, reports, and audit events before enabling paid/live model judges.

Import a text manuscript file or paste the full manuscript. Alpha limit: 1,000,000 characters.

0 / 1,000,000 characters

Model benchmark receipt

  • Track and release lane: public development, public leaderboard, private sealed, or rolling live.
  • Prompt spec, constraints, genre, rubric version, provenance, license, and contamination status.
  • Model/provider/version, decoding settings, raw output, text hash, token count, and date run.
  • Judge panel outputs, pairwise comparisons, confidence intervals, diagnostic badges, and release notes.

Author manuscript receipt

  • Manuscript, chapter, outline, scene, or revision pair.
  • Privacy mode: private agent-only, community blind, paid human-only, or public/community.
  • Requested lens: hook, pacing, dialogue realism, genre fit, KDP readiness, query readiness, market friction.
  • Credit spend, reviewer trust requirements, evidence requirements, and conflict controls.

Next executable build

The backend API now has a dry-run receipt contract for submission storage, deterministic judge outputs, and audit events. PostgreSQL-backed intake API migration scripts are ready; production DB/service activation needs the CreativeBench database and Caddy API proxy to be provisioned.