Repository navigation
fix(factory): bound routine listings and expose permit diagnostics - #6973
Merged
Merged
Conversation
Collaborator
Author
Helm Manifest DiffNo manifest changes detected across 17 application(s). |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Unfiltered routine-job MCP reads return INVALID_ARGUMENT in the connector while due-only (19 rows) and empty-kind reads succeed. The existing adapter loads and returns unbounded full rows, including payloads and summaries. Bound the response to a page (20 default, 50 maximum), omit payloads/free-form summaries, expose next_offset, and run the database read off the event loop. The connector does not expose its underlying rejection reason, so result size is the supported hypothesis, not a proven transport root cause.
Add operator-only factory_permit_diagnostics for up to twenty exact permit IDs. It uses explicit SELECT projections for reservations, session bindings/status, pending claims, stored supervision observations, DBOS status and routine leases. It excludes prompts, results, payloads, credentials and raw evidence. It does not contact guests or settle/rearm work. Persisted observations are not fresh cessation proof and never authorize recovery.
Validation: 266 local tests passed across permit supervision, factory MCP, agent MCP, routine-job behavior and catalogue compliance. Regressions cover SELECT-only access, redaction, authorization before reads, strict permit ID schema, off-loop execution, page navigation and a 100 KB free-text status. ci lint, Ruff, diff check and source ratchet passed. Independent whole-diff review found no remaining blockers after metadata string bounds were added. Real Postgres pagination passed in hosted CI. Exact-head pr-checks passed: Executed 103 out of 335 tests, all 335 passed. The HTTP smoke target passed on its automatic second attempt after a healthz ReadTimeout; no manual rerun. Image builds and repository gates passed, with zero deployment-manifest changes. BuildBuddy run.
Hosted CI reproduced an existing midnight fixture defect: the test expected today's $0.40 charge even when its paid turn was seeded twenty minutes earlier on the previous UTC day. Included the independently reviewed test-only correction to assert the paid turn's recorded date. No recovery implementation from #6972 is included.
Current head:
cb51255e237f25079d53bad8e32ffe2ba0f8ce8c.Activation and recovery remain separate. No live settings, credentials, budgets, limits, holds or permits changed. Receipt 822 and recovery draft #6972 remain held. The five live permits still need exact diagnostic evidence after a separately approved deployment before any recovery decision.