Repository navigation
docs(rfc): record the Older Sessions row-click preview as accepted - #18535
Pearcekieser wants to merge 1 commit into
Conversation
UX Review (Opus 5.5, fork) — ⏭️ skippedRevision |
First Principles Review (Opus 5.5, fork) — ⏭️ skippedRevision |
Design Review (Opus 5.5, fork) — ✅ PASSDesign-level review of Design-Verdict: PASS Docs-only RFC that records an owner decision; it removes nothing, renders nothing, and conflicts with no recorded decision. Notes
[DESIGN-REVIEWED] e3ad93a Description read[DESCRIPTION-READ] 647756b7636ecc53a845b9149b0fb6f7c15b0ddb6d2eea92bba8f5d3fe1d3646 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
GPT 6.1 Review (fork) — ✅ no blocking findingsReviewed Review detailsNo findings. |
Opus 5.5 Review (fork) — ✅ no blocking findingsReviewed |
|
note (non-blocking): Design Review, confirm the product-owner decision is posted on #15730 It is. Pearcekieser posted it on #15730 at issuecomment-6085726298, under First Principles item 1: a click or Enter on an Older Sessions row opens the read-only preview, and Resume reopens the session. That comment links this PR as the decision record. No change to the RFC is needed. |
|
note (non-blocking): Backend Tests (3.12, 2) and Coverage Gate are red from one test timeout Shard 2 failed one test, |
148d318 to
95e7b78
Compare
95e7b78 to
4bd164b
Compare
|
Rebased on main as 4bd164b, no content change. SAST (Semgrep) on 95e7b78 reported one |
4bd164b to
39b8bc5
Compare
|
Rebased on main as 39b8bc5, no content change. Main merged #18650, which appended a sentence to the same audit paragraph in |
|
note (non-blocking): GPT 6.1 Review and Opus 5.5 Review on 623ed28 both ended "review incomplete" (check-runs 123829727001 and 123829715664, finished 01:07:42Z and 01:07:32Z) without updating their PR comments, so neither reached a verdict. Their last completed reviews, on f98c5ee, found nothing, and 623ed28 carries the same 2-file diff rebased onto 0f2fbcf. I am not pushing the same tree to retrigger them; the next rebase or a maintainer re-run of the fork review lanes will cover it. The RFC decision gate (HOLD r125) still needs a maintainer call. |
|
note: rebased on main 18d6b4e as a67968e to clear another README audit-paragraph conflict, no behaviour change. Another RFC row landed in the same paragraph; this PR's row and sentence are unchanged. This push also reruns the GPT 6.1 and Opus 5.5 fork reviews, which ended "review incomplete" on 623ed28. The RFC decision gate (HOLD r125) is unchanged and still needs a maintainer call. |
|
note (non-blocking): Integration (in-process gateway) on a67968e fails on main, not on this diff. |
|
note: rebased on main 13e3ff2, no behaviour change. Main now carries #26180, which resets |
|
HOLD at 5fbce69. The RFC text is unchanged since our r125 review, and CI and every AI lane are green. Two things still block:
AI findings judged non-blocking:
— crew-pr-reviewer (r137) |
|
note: rebased on main 8742abf to clear the README audit-paragraph conflict from HOLD r137, no behaviour change. It also fills in the empty question: @buluoray, or another maintainer, could you accept this RFC? The decision it records: a click (or Enter or Space) on an Older Sessions row opens a read-only preview, and the Resume button is what reopens the session. If you accept, please name yourself and the date on the Status line and the README row, as r137 asks. Pearce made the product call on 2026-10-09. The only thing left is a maintainer's acceptance. |
Security Scope Review (fork) — 🔴 BLOCK (blocking)Which legitimate operations does the review scope was never resolved: the generate job did not finish (it timed out, failed or was cancelled), so no leg ran; its log names the step that stopped No |
|
issue (blocking): upheld, fixed in 95b9e33. The Opus review on 1f7ce10 is right. #26334 landed after the 2026-10-17 audit and added The RFC now takes what was alternative 2. The preview reads a new owner route, note: Security Scope Review on 1f7ce10 never ran its legs, because the generate job did not finish. This push gives it a fresh run. The maintainer-acceptance request above still stands. The decision being accepted, a row click previews and Resume reopens, is unchanged. Only the route the preview reads has moved. |
|
note: rebased on main 3073e60 as 8f9673f to clear another README audit-paragraph conflict, after #26380 landed in the same paragraph. The RFC text is unchanged since 95b9e33, the head Opus and GPT cleared. This push also reruns Backend Tests (3.12, 2), which failed on 15f931b in |
|
HOLD at The r137 conflict is gone (the branch merges cleanly with main) and the One blocker remains: a maintainer decision. The RFC still records Note: the sticky Security Scope Review comment still shows a BLOCK for — crew-pr-reviewer (r145) |
|
note: rebased on main 0b18d86 as 7a1f13e to clear another README conflict. This time it hit two places: the audit paragraph, and the table, where main moved the neighbouring |
|
note (non-blocking): GPT 6.1 Review, Opus 5.5 Review and Security Scope Review on 7a1f13e all ended without a verdict between 18:01 and 18:03Z. The two AI reviews report "failed before producing a verdict", and the Security Scope generate job did not finish. None of them read the diff. 7a1f13e changes only the README merge resolution; the RFC file is byte-identical to f7f910b, where both AI reviews found nothing. I am not pushing the same tree to retrigger them. The next rebase, or a maintainer re-run of the fork review lanes, will cover it. |
|
note: rebased on main 7e52ec8 as 9d5d7bb to clear another README audit-paragraph conflict. The RFC file is byte-identical to 8f9673f, the head HOLD r145 reviewed. This push also reruns the GPT 6.1, Opus 5.5 and Security Scope reviews, which could not run on 7a1f13e. The maintainer-acceptance request above still stands. |
bolichen97
left a comment
There was a problem hiding this comment.
Maintainer decision: signed at 9d5d7bb. The preview moves to a new GET /api/sessions/{key}/preview, the existing /api/sessions/{key} response is unchanged, and app tokens are not accepted on the new route. Refusing app tokens on a brand-new route takes nothing from any existing caller, so it is not added strictness.
|
note: thanks @bolichen97. 06aa168 records your acceptance as HOLD r145 asked. The Status line, the README row and the README audit sentence now name you and 2026-11-06. Nothing else changed: the RFC text is the one you signed at 9d5d7bb, apart from the Status line. |
|
note: rebased on main 7ecc6c9 as cda31ff, no behaviour change. The CI app hit "API rate limit exceeded for installation" twice on 06aa168, at 00:54 to 01:14Z and again at about 02:00Z on the re-runs, so eight checks had no verdict. This push restarts them all, including the GPT, Opus and Security Scope lanes that cannot be re-run on their own. Backend Tests (3.12, 3) also timed out in |
|
note: rebased on main 4f7b1ee as 66bd64d, no behaviour change. On cda31ff the GPT, Opus, Internal Content Scan and Security Scope lanes failed on the CI app's "API rate limit exceeded for installation" (03:08Z). Backend Tests (3.12, 6) failed on main's lazy-import pin ( |
|
note: rebased on main 594344f as 0cd8f59 to clear another README audit-paragraph conflict, no behaviour change. The RFC file is unchanged from 66bd64d. On that head the GPT, Opus and Internal Content Scan lanes failed on the CI app's "API rate limit exceeded for installation" at 05:00Z on 11-06, and Security Scope's generate job did not finish at 05:06Z. The limit keeps hitting in bursts that start near :55 and clear by about :15. These lanes start about 20 minutes after a push, so I pushed at the :00 mark. |
|
note: pushed ce96598, the same rebase as 0cd8f59 plus one README fix. Design Review's note on 0cd8f59 was right that the audit sentence gave this row two acceptance dates. It now says the row was added 2026-10-09 to record the product owner's decision, and accepted by bolichen97 on 2026-11-06. The RFC file is unchanged. The other reds on 0cd8f59 came from the CI app's "API rate limit exceeded for installation", from 03:01 to 03:12Z: Fork PR Description, Sensitive Change Review, Publish readiness signal, Dependency License Gate, and the GPT, Opus and Internal Content Scan lanes. None of them is a finding. |
|
note: rebased on main 41ba859 as 61a87a8 to clear another README audit-paragraph conflict, no behaviour change. The RFC file and this PR's README row and sentence are unchanged from 68fe543. On that head 73 checks passed and none failed. Internal Content Scan stayed in progress from 03:23Z and never reported, and this push gives it a fresh run. |
|
note: rebased on main 851ba86 as 9f16b52 to clear another README audit-paragraph conflict, no behaviour change. The RFC file and this PR's README row and sentence are unchanged from 61a87a8. On that head, Integration (in-process gateway) failed in test setup with |
Code change approver recommendation: A person should decideWhy: the decider estimates an 8% chance that this change introduces a defect, within the 12% limit for approving on its own, but it is held at tier 2: design-doc-gate: D. Docs-only PR that records, as an accepted RFC, the product decision that an Older Sessions row click opens a read-only preview and Resume reopens the session. I checked the RFC's claims about the code: in HistoryPane.tsx, activateRow dispatches resumeFromHistory on mouse-down and keyboard activation. api.sessionDetail has no consumer. api_crew_sessions_read calls /api/sessions/{key} with exclude_incognito=1, and _shape_read_messages refuses anything that isn't a list, so leaving that route unchanged and adding a new /preview route is sound. The cited helpers all exist. The front matter and status follow the RFC README's rules, and landing the decision first is the RFC-first process GOVERNANCE.md and the RFC README ask for. No trust boundary moves: the new route is owner-only and refuses app tokens. CI and AI review lanes are green; the one Security Scope BLOCK refers to an earlier head. Tier: 2, held for a person: design-doc-gate: D. An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit |
|
The design is signed off by a maintainer at 9d5d7bb. The PR now waits for a maintainer approval to merge. Review result on 6fd3c1a: no blocking findings. The PR changes docs only: one new RFC and one README index row with its audit sentence. Main does not already contain it. Every claim the RFC makes about main today matches current main: Done well: section 6 explains why reshaping Optional:
|
Problem / Motivation
Goal: Record on main, as an accepted RFC, the product owner's decision that activating an Older Sessions row opens a read-only preview and that reopening the session takes a separate Resume, so #15730 has a decision on the base branch to land against.
#15730 changes what a click on an Older Sessions row does: today it reopens the session as a tab, and after #15730 it opens a preview and leaves the session closed. Its First Principles review blocks it because no non-draft RFC on main records that product-shape change, and #12257 is an issue, not a decision. Pearcekieser, as product owner, decided on 2026-10-09 to make the change, so this document records it with
status: accepted.Why it matters
A row click is the pane's main gesture. Changing it from reopen to preview changes what every user gets from Older Sessions, and that decision belongs in a docs-only diff a maintainer can read on its own, not inside the component change.
Not a goal
What changed (motivation → approach → change)
docs/request-for-change/rfc-history-preview-not-reopen.md,status: accepted, accepted 2026-11-06 by the maintainer Bolin Chen (@bolichen97) at9d5d7bb; its Status line and README row name them and the date. It records the before/after table for each row action, the preview's states, the new owner route the preview reads (GET /api/sessions/{key}/preview, answering{key, title, messages, has_more}), the 400no_conversation_loganswer, and three rejected alternatives.GET /api/sessions/{key}keeps its bare array for every caller: since feat(instances): add crew= scope to the session read MCP tools #26334 a hub's crew transcript read calls it on a peer and refuses anything but a list, so reshaping it (the earlier design, now alternative 2) would break remote-crew reads.docs/request-for-change/README.mdgets anacceptedrow and one audit-log sentence, verified at85dcaeed2fon 2026-10-17 and re-verified at66db87b900on 2026-11-03: the row's activation (activateRow) inchat-sidebar/HistoryPane.tsxdispatchesresumeFromHistoryon click, Enter and Space, andapi.sessionDetailhas no consumer.Backwards compatibility
Compatible: documentation only; no code, config or contract changes.
Removes nothing: the diff adds one RFC file, one README row and one log sentence.
Tests
python3 scripts/docs_lint.pypasses. No code changed, so no code tests apply.Manual verification
Read sections 3 and 4 of the new RFC against the #15730 diff: each row action, preview state and response shape it lists matches a branch in
ChatSidebar.tsx,HistoryPreviewModal.tsxorapi_session_detailon that PR, and a test intest/test_session_preview.pyorwebsite/src/test/HistoryPreviewModal.test.tsx.Screenshots / video
None; docs only. The preview screenshots are on #15730.
Related Issues
Refs #12257. Unblocks #15730.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)