Website delivery verification
工程检查与模型质量分开报告。以下为本地验收记录,不是独立 CI 证明,也不是正式 Core 评测成绩。
Engineering checks are reported separately from model quality. This is a local acceptance record, not independent CI evidence or a formal Core benchmark.
Historical acceptance · 9 October 2026
- Python regression snapshot: 1164 passed, 36 skipped, with one dependency deprecation warning. Skips were not counted as passes.
- Eight Node component and proxy tests passed. TypeScript checks, the frontend build, lint and documentation consistency checks passed.
- Twelve public resource bodies matched their staged build hashes; the public visitor remained a member of the support group. Session cookies retained Secure, HttpOnly and SameSite=Strict.
- Manual browser checks covered a 375px viewport, source opening and focus restoration. A static-only preview supported the recorded answer, sources and permission comparison without an API.
Scope and limits
The October 9 record predates the subsequent showcase refinements. Its test count is not a live counter. The public historical recording uses fixed Workflow; it does not prove Dynamic Agent performance. Current formal Core Strict Task Success remains pending.
Historical evaluation data and failed or interrupted experimental runs were preserved. Public demo documents and identities are fictional.
Earlier “683 passed” homepage promotion was removed because an independent CI report was not linked. The original historical record remains in the project report.