Repository navigation
The HXI shows a build's review issues and suggestions as plain text under its message and in its failure card - #1826
Merged
Conversation
…ues and suggestions in the HXI Refs #1809. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
… in the HXI (expected failures until the build) Adds eight cases that drive the store's event handler with the IntentSurface shell open. Six fail on the base at their own assert and are marked as expected failures until the build; two controls pass on the base. Refs #1809. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
… its message and in its failure card A build_success message keeps the event's review issues and suggestions as metadata, and the failure card reads both from the report. A new component renders each list as text nodes, only when it has items, with at most ten items of at most 500 code points each, then a count of the rest. The test marker becomes a no-op. Refs #1809. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
PROGRESS.md and DECISIONS.md entries; the ledger pair regenerated. Refs #1809. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
… of record (#1809) The campaign of record ran the 14 rows on the M3 head: every mutant was killed by exactly the cases the contract names, and none survived. DECISIONS.md records the campaign. This commit changes DECISIONS.md and the two regenerated ledger files only; no source or test line changes. Refs #1809. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…ds an edited UI file (#1809) Round 1 of the review (gpt-6.1-sol, 17 min) was NOT READY on one P3: the contract's Absence Verified section claimed that no Python test reads the edited UI files, while its own list named one that does (the CapabilityApprovalView reader in test_ad1164_continue_or_ask.py). The claim now says what the list shows: one test reads an edited file and pins only what this change leaves as it is, and a second search, for a path built in pieces, finds nothing more. This commit changes the contract only; no source or test line changes. Refs #1809. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This was referenced Nov 17, 2026
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
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.
Refs #1809
What changes
A build's review lists reach the Captain. The server already sent them: a
build_successevent carriesreview_issuesandreview_suggestions(src/probos/routers/build.py:509-510), and abuild_failurereport carries both (src/probos/cognitive/builder.py:177-178,:201-202). The HXI showed the success message with a suggestion count and the failure card with the category, the title, the failed tests and the raw error, so the Captain never saw the review's own text. Now:build_successhandler passes the two lists as an optionalbuildReviewon the chat message (and onaddChatMessage's meta, besidebuildFailureReport), andIntentSurfacerenders them directly after the message bubble.review_suggestions, and the failure card shows both lists after the failed tests and before the raw error.ui/src/components/chat/BuildReviewLists.tsx(new, 70 lines) takes twounknownvalues, checks their shape, applies the limits and renders two titled lists ("Review issues" and "Review suggestions"). A non-array shows nothing, non-string and empty items are skipped, and a list with no items renders nothing, not even its title.lias a JSX text expression, outside the ReactMarkdown subtree, so model text is never read as markup. At most 10 items show per list, each cut at 500 code points (Array.from, so a surrogate pair is never split) with an ellipsis (U+2026), and "and k more" counts the string items not shown.The message text, the server, the event payloads and the stored chat history are unchanged. Nothing is persisted, so the lists do not survive a reload. UI only: no Python, config, flag or schema change.
One new test file,
ui/src/__tests__/BuildReview.bf978.test.tsx(160 lines, 8 cases: 6 that fail on the base and 2 controls), uses no timer. The rest of the diff is the contract (prompts/build-review-lists.md, 728 lines),DECISIONS.md(+9),PROGRESS.md(+5) and the regenerated ledger pair.Six commits: the contract; M1 (the new tests, six of them marked as expected failures until built); M2 (the types, the store, the component,
IntentSurface, and the marker becoming a plainit); M3 (the trackers and the ledger pair); R1 (the mutation campaign's record inDECISIONS.md); and R2 (one claim in the contract, found by review).Review and the broad gate
Every review was a Diff Reviewer on
gpt-6.1-sol. The Architect drafted the contract; the Builder applied M1 and M2, each blob checked against a tool-applied oracle; the contract, M3, R1, R2 and the ledger pair were applied by tool.Round 1 (the whole candidate, 17 min) found the UI sound in every consumer it named (the chat thread, the plain-text rendering, both limits, empty values,
localStorage, every otheraddChatMessagecaller); it ran all 14 mutation rows and got the contract's killers. It was NOT READY on one P3 in the contract: the Absence Verified section claimed that no Python test reads the edited UI files, while its own list showed one that does (tests/test_ad1164_continue_or_ask.py:1082readsui/src/store/types.tsand pins only theCapabilityApprovalViewdeclaration, which this change does not touch). I confirmed it with the contract's own search and a wider one (every Python file that namestypes.ts,useStoreorIntentSurface): that test is the only reader of a real edited file. R2 (the contract only, +3/-3) states that, and records the wider search. Round 2 (the delta, 9 min) was READY: the finding was gone and nothing new was found; the reviewer enumerated 1,940 Python files undertests/andscripts/for readers of the three edited files and found the same single one.Broad gate, on
3441afc7d91359b0b3541e831d07696bb597c14f: one run in the canonical wrapper at--workers 3:54111 passed, 184 skipped, 246 warnings, 7 subtests passed in 1:34:14, exit 0 (started 13:55Z). The receipt (logs/gates/build-review-lists.receipt.json) records 54,295 tests with 0 failures and 0 errors, binds this head and its treef5ad12fc6a960572e1006887260e167036b5eb2e, and validated four times (three after the run, once before the push). The live-data manifest was identical before and after the run (54,107 files, none added, removed or changed). This change touches no Python file, so the run is the guard that the UI-only diff leaves every Python consumer alone; the UI itself was checked by the focused vitest sets,tsc -bandnpm run buildbelow, and by CI's fullui-testsjob.Evidence
BF978_RUNXFAIL=1exactly S1, S3, S4, S5, S6 and F1 fail, each at its list assertion (expected null not to be null), while the controls S2 and F2 pass.IntentSurface.*files,useStore.test,BuildQueueApprove.bf937,AgentProfilePanel.threadKeyed,App.bf710andProfileChatTab.ad1291): 170 passed in 16 files on the base, 178 passed in 17 files at the head in both modes.npx tsc -bandnpm run buildexit 0, the gate wrapper's preflight-only exits 0, and the ledger test passes.3441afc7d913(python-tests-shard1, 2 and 3,python-testsandui-tests), 16:10Z. The first attempt failed two timing tests inpython-tests-shard (3): the AD-1192 1000-row test hit its 300 s budget (already tracked by Follow-up: residual findings from the #1375 / #1127 / #1054 wave #1414 and Follow-up: residual findings from the #1239 / #1515 / #1511 wave #1523) and an identity.db unit test missed its own 0.2 s test bound. This change touches no Python file, and both tests passed in the canonical gate on this head (123.18 s and 0.33 s). The rerun of the failed jobs, with no source change, passed.What this pull request does not do
build.py,builder.py), the event payloads, the stored chat history or the type ofBuildFailureReport.review_issues.cognitive/builder.py:2538-2550), which stay AD-1359 follow-up: surface crew size suggestions and readings; pin the dev-side size rules #1769's review point.