Skip to content

spec(epic): record FIX-1794 per-conversation task board as D6 (FIX-1786) - #2831

Merged
jhoffner merged 5 commits into
mainfrom
spec/FIX-1786-amend-4
Oct 6, 2026
Merged

jhoffner merged 5 commits into
mainfrom
spec/FIX-1786-amend-4

Conversation

@jhoffner

@jhoffner jhoffner commented Oct 6, 2026 •

Copy link
Copy Markdown
Collaborator

Follow-up to #2795 (the FIX-1786 epic spec), amended by #2806, #2807, #2813, #2822 and #2824, all merged. Conditional on #2828 (FIX-1794's spec): it merges together with that gate. If FIX-1794's D1 changes in review, this PR is updated to match before merge.

The problem

A user can have two conversations open, and each one hands tasks to workers on other flows. Today the only store that crosses a flow is the user's own scope. That store is one list for every conversation the user has, so one conversation can take and run another's tasks. FIX-1794's POC shows it (U1). A filter in Workforce stops the take, but the conversation still lists and waits on the other's tasks (E1). The fix changes the shared task board, which is Layer 1. Per ER-22 and ER-24, a Layer 1 change binds the set only after the epic records it.

What changes

  • New decision D6. A board keeps its tasks at its owner's user scope, in a partition only its own conversation reaches. Every board in the chain has this shape. The engine does not change.
  • D3 now has four Layer 1 changes, not three. ER-22 and the box (SPEC.md, end-state.svg, d1-the-set.svg) say four.
  • ER-9 states the shape and no longer leaves it as FIX-1794's open question. FIX-1792 is a new consumer of ER-9 in the ownership matrix. ER-24 records D6 as met.
  • FIX-1794 comes before FIX-1792. FIX-1792 moves each session board onto FIX-1794's conversation board. The graph gets that edge, and PLAN step 4 now waits on both. The path figure already put FIX-1792's build after FIX-1794's, so the critical path does not change.
  • One exception in step 4. FIX-1794's first PR, the partitioned board, touches only orchestration on its own fixtures, so it may start once this merges, without waiting for FIX-1791. This matches FIX-1794's PLAN.
  • FIX-1788 post-merge amendment: the worker-session lookup matches on the key set (specs/issues/FIX-1788/ S5a, BR-14a, V4, EVOLUTION). A criteria key the lookup does not name matches only sessions that do not carry that key. Without this rule, a narrower lookup lands in a wider session: plain talk with a worker opens its delegate or task session, and a coordinator's post to a delegate lands in a task session (found in spec(FIX-1794): tasks go down the owner's chain and run as the owner #2828's round-2 review). The rule is stated once in FIX-1788, and FIX-1791, FIX-1793 and FIX-1794 cite it. No FIX-1788 build PR is open yet.
  • The concept, the board seam in PLAN, D5, and a line in DECISIONS → How it got here.

No epic rule talks about chain depth, so FIX-1794's limit of five (its D2) is not recorded here.

Who owns what: ER-9, a chain runs as the owner of the work, is decided by FIX-1794, consumed by FIX-1793 and FIX-1792, and checked by the closure

The ownership matrix: ER-9 now has FIX-1792 as a consumer.

Reviewers · look here

What's asked of you

Should one user's conversations keep separate task lists, if that means a change to the shared task board? I recommend yes, and you decide it at #2828. This PR only records that answer for the whole set. If you say no there, the alternative is a change to the engine, and this PR is rewritten before merge. Until it merges, main shows three Layer 1 changes and FIX-1794's question as open, and FIX-1794's first PR can't start.

After merge (coordinator): in Linear, change FIX-1794 → FIX-1792 from related to blocks.

Links: FIX-1786 · original review #2795 · amendments #2806, #2807, #2813, #2822, #2824 · FIX-1794's spec #2828 · set at specs/epics/FIX-1786/

How to review this: the epic-spec reviewer contract

This is an epic-spec: the shared objective and cross-cutting decisions for a set of
issues, in a retained document set. It is not any one issue's implementation design.

In scope to challenge:

  • The objective and the goal in SPEC.md — is this body of work worth doing, and is the
    goal the real need at full size, or a smaller one the set could meet while the need stays
    unmet? Would the closure issue's check fail if the goal were not met?
  • Whether the set overbuilds. Each issue can earn its place while the whole is too much.
    That question can only be asked here.
  • A cross-cutting decision in DECISIONS.md — shared surface, naming, sequencing, contracts.
  • A rule in BUSINESS-RULES.md with no owner, or two.
  • A missing issue the objective implies, or one in the set that doesn't serve it.
  • Shared promises and publication ownership in DOCS.md; missing or misleading
    predecessor treatment in conditional EVOLUTION.md.

Out of scope — owned by the individual issue specs:

  • Any single issue's approach, architecture, file layout, or test plan.
  • Anything that touches exactly one issue. It belongs on that issue's spec PR.
  • The figures, at the pixel level, and routine status movement since the dated snapshot.
  • POC production polish. End-state experiments under the owning spec's poc/
    show shape and scoping, not shippable implementation. Directional evidence, experimental
    isolation and absence of secrets/generated dependencies remain in scope.

Feedback in the second list is routed to the issue it concerns as an implementer note, not
folded in here.

Required repository checks, approvals and review-thread policy still govern merge.
Optional comments are triaged, not a reason to grind to zero or restart the design review.

🤖 Generated with Claude Code

https://claude.ai/code/session_01XMg2dz8sMEdyzCQTRsaMTV

claude added 2 commits October 6, 2026 22:04
…1786)

FIX-1794's D1 keeps a board's tasks at its owner's user scope, one
partition per conversation. It changes the task board (Layer 1), so per
ER-22 and ER-24 it binds the set only once the epic records it. D6
records it; D3 counts four Layer 1 changes; ER-9 states the shape and
gains FIX-1792 as a consumer; PLAN's FIX-1794 row and board seam, the
box, the concept and the ownership matrix follow.

Conditional on #2828's approval: if its D1 changes in review, this is
updated to match before merge.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XMg2dz8sMEdyzCQTRsaMTV
…version (FIX-1786)

D6 gives every board in the chain one shape, so FIX-1792 moves each
session board onto FIX-1794's conversation board. The graph gains that
edge, and PLAN's rows and step 4 put FIX-1794 before FIX-1792. The path
figure already draws FIX-1792's build after FIX-1794's; the critical
path does not move.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XMg2dz8sMEdyzCQTRsaMTV
@vercel

vercel Bot commented Oct 6, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

3 Skipped Deployments
Project Deployment Actions Updated
flow-state-dev-docs Ignored Ignored Oct 6, 2026 10:24pm UTC
flow-state-dev-kitchen-sink Ignored Ignored Oct 6, 2026 10:24pm UTC
flow-state-dev-ui Ignored Ignored Preview Oct 6, 2026 10:24pm UTC

Request Review

@linear-code

linear-code Bot commented Oct 6, 2026 •

Copy link
Copy Markdown

FIX-1786

FIX-1794

…-1786)

Mirrors FIX-1794 PLAN at a0b1226: P1 touches only orchestration on its own fixtures.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XMg2dz8sMEdyzCQTRsaMTV

jhoffner commented Oct 6, 2026

Copy link
Copy Markdown
Collaborator Author

from: fsd-head-of-engineering / session: em-fix-1786 / kind: ci-standing-down

Test (3/3) failed at 636e905. The failing test is @flow-state-dev/memory → test/capture-evaluator.test.ts › "a delayed skip racing an evaluator failure never marks the failed turn read". The assertion at :427 expected the last judged state to contain turn C, but it held only turn D.

This failure is not caused by this PR:

  • The diff touches only 8 files, all under specs/epics/FIX-1786/. No code changes.
  • The same job passed at the previous head, 2a36df0.
  • The test exercises timing between two overlapping captures, which no spec file can influence.

No fix exists yet; I found no open PR touching this test. It may also point at a real intermittent bug in memory capture, where a late skip marks a failed turn as read. That is outside this PR, so I am raising it separately instead of widening this one.

I am re-running the failed job once.


Generated by Claude Code

Post-merge amendment carried on the epic's amend-4. A criteria key the
lookup does not name matches only sessions that do not carry it, so a
narrower lookup never lands in a wider session (plain talk in a delegate
or task session; a delegate post in a task session). Stated once in S5a
and BR-14a; FIX-1791, FIX-1793 and FIX-1794 cite it. V4 gains the S12 case.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XMg2dz8sMEdyzCQTRsaMTV
jhoffner pushed a commit that referenced this pull request Oct 6, 2026
A post to a delegate never lands in a task session: S9 and BR-19 cite
FIX-1788 S5a's key-set matching as amended in #2831, with a V7 case.
A split parent's settle is owed on the row like a notice, so a failed
turn can't strand it (S8, BR-32, V6, conditional on Q). The wake marker
is written with the add; BR-10a states the next-touch limit; "only an
add starts a run" covers marker retries; the breadth cap sits beside
D2 and travels with the split; the key-prefix test is a new file.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XMg2dz8sMEdyzCQTRsaMTV
@jhoffner
jhoffner marked this pull request as ready for review October 6, 2026 22:19
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 6, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-06T22:23:26.411106Z bebe2ff Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@cursor cursor Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Simplify review

Problem this PR solves: Cross-flow task chains need boards that stay isolated per conversation at the owner's user scope. FIX-1794's gate POC ruled out the lighter options (one unpartitioned user ledger, Workforce-only claim filtering, reply-settle in the task session). Recording that as epic D6 and threading it through ER-9, D3, and the plan is the right binding move — same pattern as Q1 after FIX-1789.

Verdict on approach: The partitioned ledger is the cleanest fix among the rejected horns; I would not ask you to reopen that fork. What did grow faster than necessary is how many places say the same thing, not the underlying model.

What looks like the right level of complexity

  • D6 as a full decision card with POC citations and explicit "instead of" list — implementers need one canonical card.
  • ER-24 + ownership matrix update (FIX-1792 consumes ER-9) — keeps the set coherent.
  • FIX-1788 S5a key-set match — complements D6 (session lookup vs ledger partition). Bundling it here is defensible, but the PR reads like two amendments unless you tie them in one place.

Simpler shapes worth considering (non-blocking)

  1. Single-source mechanism — ER-9, D3 item (4), CONCEPT, and D6 all spell out partition + server-named keys. Consider ER-9 as outcome + link to D6 only (closer to pre-amendment ER-9), and let D6 own mechanism prose. Same information, fewer places to drift.
  2. Layer 1 count — Bumping D3 to four because orchestration gains a partition hook is consistent with ER-22, but it's a real tax (figures, sign-off, D1 cost row). If you want a lighter epic surface, an alternative is: keep D3 at three engine-facing changes and treat conversation partitioning as FIX-1794's deliverable bound by ER-9 without a fourth Layer 1 slot. I'm not sure that's better for readers who scan "Layer 1" as "anything Workforce can't fake" — flagging as a judgment call, not a required change.
  3. Decision tree — D6 is both a peer of D3 in the mermaid and item (4) inside D3. A D3 --> D6 edge (or D6 only under D3 with Q1-style epic anchor) would reduce dual hierarchy.
  4. Links — D6 points at specs/issues/FIX-1794/... paths that are not in main yet; until #2828 merges, prefer PR/issue links only (you already have #2828) to avoid broken relative links in the retained spec.
  5. Evolution hygiene — The "leaves how to FIX-1794's spec" bullet now ends with "It came back as D6"; one pass to drop the "leaves how" half would read cleaner. Epic EVOLUTION.md still looks one gate behind DECISIONS on D6 (optional follow-up).

Skipped (needs FIX-1794 implementation spec, not this amend)

  • Whether partition maps to existing defineTaskCollection instance keys vs a new orchestration primitive.
  • Store-side indexing for listSessions + key-set filter (perf smell on hot ensureWorkerSession, not wrong for v1).

Approval: Directionally sound; POC-backed; plan sequencing (1794 → 1792, early D6 PR exception) is practical. Nits above are deduplication and navigation, not a rethink of D6.

Open in Web View Automation 

Sent by Cursor Automation: Review For Best Approach

Comment thread specs/epics/FIX-1786/BUSINESS-RULES.md Resolved
Comment thread specs/epics/FIX-1786/DECISIONS.md Resolved
Comment thread specs/epics/FIX-1786/DECISIONS.md Resolved
Comment thread specs/epics/FIX-1786/DECISIONS.md Resolved
Comment thread specs/epics/FIX-1786/PLAN.md Resolved
Comment thread specs/issues/FIX-1788/PLAN.md Resolved

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: bebe2ff877

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread specs/epics/FIX-1786/DECISIONS.md Resolved
Comment thread specs/epics/FIX-1786/PLAN.md Resolved
Comment thread specs/issues/FIX-1788/PLAN.md Resolved
…y-set check (FIX-1786)

- PLAN step 4: FIX-1793's board-consuming slice merges only after
  FIX-1794's first PR (the partition its workstream boards live on).
- FIX-1788 V4: the strict-subset case, so a coordinator lookup never
  lands in a task session carrying its keys plus taskId.
- FIX-1788 EVOLUTION names the session layer beside D6's ledger layer.
- DECISIONS: the ER-9 history bullet reads as resolved.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XMg2dz8sMEdyzCQTRsaMTV
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants