[buttercode101]

Unified product

AgentScope remains the user-facing, local-first supervision product. Habitat is integrated as its evidence, claim-verification, policy, action-ledger, and portable-proof runtime.

Implemented

• Vendored all 28 Habitat runtime Python modules from upstream commit b823836d826a73f1edcea3ff0c5b7ba7399a6219.
• Imported Habitat's regression suite, CLI examples, schema, proof tooling, and public proof fixtures.
• Exposed the habitat CLI from the unified AgentScope Python distribution.
• Added a read-only SQLite bridge to AgentScope's local dashboard generator.
• Dashboard now displays Habitat health, job counts, active signals, stored claim verdict counts, and action-ledger integrity.
• Missing or unreadable evidence is UNKNOWN; failed ledger integrity is surfaced as FAIL.
• Added optional sources.habitat_db config.
• Updated product documentation and landing-page messaging to explain the evidence layer.

Trust boundaries

• AgentScope opens Habitat's DB in SQLite read-only mode.
• The dashboard does not execute jobs, mutate claims, or acknowledge signals.
• Stored claim verdicts are shown as stored state, not freshly re-verified.
• Proof validity is not presented as proof of external truth.
• Both original repositories and their main branches have been preserved.

Verification

• Unified CI: 136 passed.
• Vercel preview: landing page and demo return HTTP 200; evidence-layer messaging verified on the preview.
• Latest Vercel build check hit the team's deployment build-rate limit. This did not affect GitHub CI. The latest small documentation/runtime-requirement change is therefore not yet reflected in the preview.
• Production promotion is intentionally not claimed or performed.

[vercel]

Deployment failed for project agent-supervision with the following error:

Resource is limited - try again in 24 hours (more than 100, code: "api-deployments-free-per-day").

Learn More: https://vercel.com/butters-projects-ad226a08?upgradeToPro=build-rate-limit