Why: A rejected checkpoint did not stop a later paid request, and a history-only fix could miss JSON errors or other active runs.
Evidence: Source and JSON rejection fixtures failed before the instance-level guard, then passed with fail-closed and unrelated-tool checks; independent replay matched both real failed databases.
Impact: Read-only event correlation blocks future provider requests after a checkpoint rejection; original sample failure and stop-rule deviation remain recorded.
Refs: Agent-System-Backend#193; Agent-System-Dev#512
Why: An approved one-shot sample continued to a paid model call after a nonliteral quote was rejected, contrary to its stop rule.
Evidence: A zero-paid provider fixture failed before the guard and passed afterward; the original 14231-token deviation remains recorded.
Impact: Future controlled evaluations refuse subsequent provider requests after a checkpoint tool error; this does not reclassify the failed sample.
Refs: Agent-System-Backend#193; Agent-System-Dev#512
Why: The prior one-shot A01 result failed and its resource approval is exhausted.
Evidence: Follow-up authorization tests failed before the change and pass after; full Backend suite passes.
Impact: Only live evaluation harness gains a separate one-shot result and 30000-token run budget; production validation remains strict.
Refs: Agent-System-Backend#193; Agent-System-Dev#512
Why: Upstream now reserves model and mock settings for system authority.
Evidence: Complete merged Backend suite and live resource/cap guards pass.
Impact: Fixture-only startup configuration, same effective route and profile; no further paid calls.
Refs: Upstream scoped settings contract; U2/U3 evaluation
Why: The user approved one additional A01 sample after the original six-case batch failed its quality gate.
Evidence: Live fixture tests preserve the production output cap and reject paid requests when a third actual run exists.
Impact: Evaluation-only case selection and resource guard; no production behavior changes, no automatic replay.
Refs: User approval 2026-10-13; reported token target 60000
Why: A reconciled task must not retain the old queue or running next step after becoming interrupted.
Evidence: Owned queue recovery failed before the fix; RunMgr, actual App startup and evaluation output-cap checks now pass.
Impact: Recovery uses the stopped-task next step, and the mechanism fixture quotes the actual acceptance source text.
Refs: Approved recovery contract; independent terminal-state review
Why: Authentic quotes do not establish an unstated approval, and the evaluation wrapper must honor tighter production output limits.
Evidence: Approval-overstatement and 780-token-cap regressions failed first; complete Backend suite and fixture cap checks now pass.
Impact: Render material records from trusted labels and validated quotations, reserve interpretation for explicit finding kinds, bump recipe identity and cap the fixture without overriding smaller limits.
Refs: Approved R05; independent model assessment; original failed samples preserved
Why: The first real-model sample batched discoveries, overcalled scope differences and exposed process chatter as a terminal conclusion.
Evidence: Preserve the failed A01 sample; process-text regression failed before the fix, then complete Backend and focused runtime checks passed.
Impact: Bump recipe identity, qualify quantities and approvals, exclude Process parts from canonical results and capture independent historical evaluation snapshots.
Refs: Approved U2/U3 plan; independent model review and native SELF receipts
Why: Text with short attachments previously completed without starting the intended background material task.
Evidence: SELF real HTTP/browser reproduction shows the absent second task; runtime, manager and server regressions pass with the existing optional port connected.
Impact: Preserves dispatch timing and makes deferred intake observable; comparison records the baseline gap without rerunning a sample.
Refs: fast-slow-u2-u3 A02 and reading acceptance
Why: Opening a processed source must not show changed current bytes as the quoted version.
Evidence: SELF TaskSourceContent regression fails before and passes after expected-hash enforcement; artifact regressions and live compilation pass.
Impact: Optional content hash narrows the existing authorized artifact read; controlled fixtures select a genuinely provided quotation.
Refs: fast-slow-u2-u3 source-version acceptance
Why: Baseline and candidate must exercise their actual default prompts with identical route and resource limits.
Evidence: SELF live-tag compilation passes.
Impact: Evaluation harness only; normal automatic notice quality remains outside the controlled comparison.
Refs: fast-slow-u2-u3 Task 7
Why: The live harness must use the actual client constructor and preserve MiniMax reasoning separation.
Evidence: SELF live-tag compilation passes without provider calls.
Impact: Test harness only; no production route changes.
Refs: fast-slow-u2-u3 Task 7
Why: Real source-backed discoveries should reach the same task card before the slow channel completes.
Evidence: SELF receipts record runtime and manager checkpoint red-green tests, scoped source/version checks, privacy filtering, quota corrections, and SQLite/PG regressions; fixed composition and live comparison follow.
Impact: Adds a local material-only checkpoint sink with no summary model calls, uses version witnesses for source commits, and preserves existing reading and notification paths.
Refs: fast-slow-u2-u3 Tasks 6 and 7; test-only live HTTP harness bounds shared provider requests without approving HITL
weichen