v0.1.0-alpha.1 — Research Preview

The first research preview of Human Value Allocation turns a text conversation into a report of visible human contributions, with source quotes and immediate assistant followup.

• Six contribution categories with a public, editable evaluator.
• Local HTML and JSON reports; invalid evidence and failed calls remain unresolved.
• A sanitized agent-eval-v2 engine subset, explicit model configuration, and no bundled credentials or production data.
• Offline illustrative demo plus an API-backed analysis command.

Scope: descriptive contribution analysis. Causal attribution, token-equivalent value and task-quality gains are future research; this rubric is not yet calibrated.

Try the example report · Quick start · Validation record

Validation: 17 tests passed in the checkout and an isolated wheel installation; four synthetic human turns completed live analysis with verified quotes. Gitleaks found no secrets in the new history or built distributions. The demo uses authored illustrative annotations and is not a performance benchmark.