Repository navigation
docs(rfc): Mate gets one landing, then the ordinary landing applies - #18903
Conversation
|
@buluoray this amends your rfc-crewmate-guides-and-mate "When it opens": a never-chatted Mate stays the first-visit landing, but only while the user has no other crewmate they chatted with (last_chat_ts) or last opened. Today a returning user is moved onto Mate on every bare visit. Could you accept (or adjust) it? The implementation PR follows and links here. |
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — 🟡 CONCERNSDesign-level review of Design-Verdict: CONCERNS The "Seen" rule (marker claimed or thread has a row) misses a Mate whose welcome is never owed. That Mate stays empty and unclaimed, so it pins the user on every visit, as today. Watch
Suggestions
[DESIGN-REVIEWED] 9d23c57 Description read[DESCRIPTION-READ] 161524e8d44d98f7e5594cf9a2e679ff94c1c291647a14ec351c5792cf3c36f6 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 |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
There was a problem hiding this comment.
Requesting one change before I accept this as the author of rfc-crewmate-guides-and-mate. Reviewed at 79a58e3291d37604b723ed01ffe6a206e96b79e0.
The rule I want: Mate gets one landing, for fresh and existing users alike. Once Mate has sent its welcome, it counts as seen, and the ordinary landing applies. The original decision put Mate first "on fresh and existing installs alike" on purpose, so existing users get introduced to it. A rule where a returning user is never shown Mate unless they go looking drops that.
As far as I can read main, that is already what happens when the welcome lands, which the Problem section should say:
- The welcome is a real assistant row in Mate's thread (
dashboard/mate_welcome.py, "The greeting itself is the crewmate's own assistant row"). has_dm_messagecounts it, including a live row that has not flushed yet (dashboard/handlers/members.py:648-653).- So
pendingMate(website/src/lib/assistantMember.ts:22-27) stops answering Mate after the welcome, andresolveMateLandingyields tolastChattedMember/ the remembered crewmate on the next bare visit.
The case that does pin a user on Mate every visit is a welcome that never lands. The once-only marker is claimed before dispatch (claim_greeting_marker), so a failed welcome turn (signed-out CLI, no backend) leaves the thread empty for good, and pendingMate keeps answering Mate on every bare visit with no welcome coming.
What I'd accept this RFC as:
| Mate's welcome | Lands on |
|---|---|
| not sent yet | Mate (one introduction, fresh and existing users) |
| sent | the ordinary landing (last-chatted, then remembered crewmate) |
| attempted and failed | the ordinary landing (counts as seen) |
- "Seen" = the welcome marker has been claimed, or Mate's thread has any row. That is server-side state, so it holds across browsers and restarts with no new client marker.
- If that state cannot be read, fall back to the ordinary landing, so a read failure never pins the user on Mate.
?member=still wins,defaultstill does not count, and the user can always open Mate from the roster.- §4 should list "Mate yields once its welcome is sent or attempted" and why it is chosen over both "Mate first until the user replies" and "never land on Mate for a returning user".
- §5's tests: an existing user lands on Mate once; after the welcome row exists, the next bare visit lands on their last-chatted crewmate; with a claimed marker and an empty thread, the bare visit also lands on the last-chatted crewmate.
Once the document says this, I'll accept it here and the status can flip to accepted.
79a58e3 to
db3ef22
Compare
Amends rfc-crewmate-guides-and-mate "When it opens" in its own document, on the rule buluoray asked for in the #18903 review: Mate lands once for fresh and existing users; once its welcome marker is claimed or its thread has a row, the ordinary landing applies. Draft; acceptance asked of buluoray.
db3ef22 to
9d23c57
Compare
|
@buluoray thanks, updated to your rule at 9d23c57 (one commit, docs-lint passes). The file is renamed to Point by point:
#18904 is being reworked to match: the "another crewmate was chatted with" rule is dropped. |
buluoray
left a comment
There was a problem hiding this comment.
Thanks, this is the rule I asked for, and §1's diagnosis of the failed-welcome pin matches main. One gap before I accept it, the one the Design lane raised, which I checked at origin/main:
A Mate whose welcome is never owed stays "not sent yet" forever. maybe_start_first_greeting returns NOT_OWED before it claims the marker (mate_welcome.py:208-209), and welcome_owed is only written by first_crewmate._owe_first_welcome when Mate is created fresh (first_crewmate.py:180-184, whose docstring says a write failure "only costs the welcome"). So a Mate that predates the owed record, or whose record failed to write, never claims the marker and never gets a row. Under §2 that Mate reads as "not sent yet", and pendingMate answers Mate on every bare visit: the same pin §1 removes, through a different path. NO_THREAD and NOT_CREWMATE leave the same unclaimed, empty state.
Ask:
- §2: "not sent yet" means a welcome is still coming: owed, unclaimed, and the thread is empty. Anything else (not owed, refused, claimed, or a row exists) counts as seen and gets the ordinary landing. The roster row carries the owed state beside the claimed state.
- §5: add a test for an empty Mate with no owed record and no marker, which lands on the ordinary landing.
Not blocking: §4 could name "release the marker when the welcome turn fails, so it is retried" and say why it was not chosen.
Once that lands I'll accept the RFC.
buluoray
left a comment
There was a problem hiding this comment.
Approving, and this supersedes my request-changes review on this head. As the author of rfc-crewmate-guides-and-mate, I accept this amendment: Mate gets one landing, and once its welcome is sent (or attempted and failed) the ordinary landing applies. Please flip status to accepted when you merge.
On the "welcome never owed" case I raised: Mate is the single built-in assistant and always sends its welcome, so that state is not one this rule needs to cover. No change needed to §2 or §5 for it.
…r a failed welcome Mate's landing (#18406) answered Mate while its thread was empty. The once-only welcome marker is claimed before the welcome is dispatched, so a failed welcome left the thread empty for good and every bare visit landed on Mate. The roster row now reports whether Mate's marker is claimed (welcome_claimed), and resolveMateLanding yields once the marker is claimed or the thread has a row; an unknown state uses the ordinary landing (rfc-mate-landing-once, #18903).
…r a failed welcome Mate's landing (#18406) answered Mate while its thread was empty. The once-only welcome marker is claimed before the welcome is dispatched, so a failed welcome left the thread empty for good and every bare visit landed on Mate. The roster row now reports whether Mate's marker is claimed (welcome_claimed), and resolveMateLanding yields once the marker is claimed or the thread has a row; an unknown state uses the ordinary landing (rfc-mate-landing-once, #18903).
…ion (#19064) buluoray approved #18903 and asked for the status to be flipped. Folds in the one gap that review raised: a Mate whose welcome was never owed is refused before the marker claim, so "still coming" now means owed, unclaimed and no row; anything else gets the ordinary landing. Adds the "release the marker on failure" alternative to section 4. Co-authored-by: Joe Guo <zejiangg@amazon.com>
…r a failed welcome Mate's landing (#18406) answered Mate while its thread was empty. The once-only welcome marker is claimed before the welcome is dispatched, so a failed welcome left the thread empty for good and every bare visit landed on Mate. The roster row now reports whether Mate's marker is claimed (welcome_claimed), and resolveMateLanding yields once the marker is claimed or the thread has a row; an unknown state uses the ordinary landing (rfc-mate-landing-once, #18903).
…r a failed welcome Mate's landing (#18406) answered Mate while its thread was empty. The once-only welcome marker is claimed before the welcome is dispatched, so a failed welcome left the thread empty for good and every bare visit landed on Mate. The roster row now reports whether Mate's marker is claimed (welcome_claimed), and resolveMateLanding yields once the marker is claimed or the thread has a row; an unknown state uses the ordinary landing (rfc-mate-landing-once, #18903).
…r a failed welcome (#18904) Mate's landing (#18406) answered Mate while its thread was empty. The once-only welcome marker is claimed before the welcome is dispatched, so a failed welcome left the thread empty for good and every bare visit landed on Mate. The roster row now reports whether Mate's marker is claimed (welcome_claimed), and resolveMateLanding yields once the marker is claimed or the thread has a row; an unknown state uses the ordinary landing (rfc-mate-landing-once, #18903). Co-authored-by: Joe Guo <zejiangg@amazon.com>
Problem / Motivation
Goal: Record the decision that Mate gets one landing on the Crewmates page, for fresh and existing users alike, and then the ordinary landing applies.
A user was moved onto Mate on every bare Crewmates visit. When Mate's welcome lands this already ends (the welcome row sets
has_dm_message). It never ends when the welcome fails:claim_greeting_markeris claimed before the welcome is dispatched, a failed turn leaves the thread empty, andpendingMateanswers Mate on every visit.Why it matters
A user whose welcome failed (signed-out CLI, no backend) is pinned on an empty Mate for good and never reaches the crewmate they last chatted with.
Not a goal
rename_self, or the once-per-visit rule below md.What changed (motivation → approach → change)
Edited 2026-10-11 to the rule buluoray asked for in the review (the first draft yielded whenever another crewmate was chatted with; that is now §4's rejected alternative).
docs/request-for-change/rfc-mate-landing-once.md(draft): Mate lands once; once its welcome marker is claimed or its thread has a row, the ordinary landing applies; an unreadable state falls back to the ordinary landing;?member=wins,defaultdoes not count, Mate stays on the roster. §4 picks "yields once its welcome is sent or attempted" over "first until the user replies" and "never for a returning user". §5 lists the tests. Claims checked at73d44b420b.docs/request-for-change/README.md: index row and audit line.Backwards compatibility
Compatible: docs only.
Removes nothing: adds one RFC document and one index row.
Tests
N/A — docs only.
scripts/docs-lint.shpasses.Manual verification
N/A — docs only. The code claims (
claim_greeting_markerbefore dispatch inmaybe_start_first_greeting;has_dm_messagecounts live rows;pendingMatereads only rows) were read at73d44b420b.Related Issues
Amends
rfc-crewmate-guides-and-mate.md(#18326, implemented in #18406). Implementation: #18904.Checklist
Contribution License Agreement
Contributed under the project's Apache-2.0 license.