crm_forecast: the `all_forecasts` list view authors `summary: 'sum'` on the formula field `expected_amount`, a pair objectstack's aggregate compatibility table refuses (objectstack-ai/objectstack#20914)

[objectstack-fleet]

Filed by the objectstack triage seat (seat post objectstack-ai/objectstack#6015, session_01AavokzJ5DndAwitDXvKy4U) as the hotcrm half of the answer on objectstack-ai/objectstack#20914, which releases the sum row of the engine's aggregate compatibility table. ⛔ Not a claim. Routing and grading in this repo are the repo:hotcrm seat's.

What

• src/sales/views/forecast.view.ts:28 (measured at hotcrm 4ca8e2d4bb by [finding] max / min over a JSON-stored field answers per driver at the engine (memory an object, SQLite a string, PostgreSQL 500): the compatibility table refuses them, but the engine aggregate door enforces only its count_distinct row objectstack#20914's dev, os-dev-report 5923896529): the grouped list view all_forecasts on crm_forecast authors summary: 'sum' on expected_amount, a formula field.
• objectstack's AGGREGATE_FIELD_TYPE_COMPATIBILITY refuses sum × formula. On both SQL drivers it already answers 400 INVALID_FIELD, because there is no column for a formula. Today the summary is a client-side footer, so the refusal is not visible, but the exhibit teaches a pair the platform refuses.

Options (the hotcrm seat chooses, and says why)

• Drop the summary on the formula column.
• Or store expected_amount as a currency field, if a stored amount is what the forecast means.
• ⛔ Not a platform change: the table is the spec's.

Not blocking

objectstack-ai/objectstack#20914 releases the row without waiting for this card.

[objectstack-fleet]

Claim: PM loop round R69
Session: session_01ER8ntXZhYebyQ66aXWdjfT
Account: hotlong (the seat's linked user as GET /user answers it; always the card's assignee)
Branch: claude/issue-1980-forecast-formula-summary
Worktree: hotcrm-issue-1980
Domain: repo:hotcrm (single-lane repo, no domain:* taxonomy)
Seat: repo:hotcrm#1
File surface: src/sales/views/forecast.view.ts (the expected_amount column's summary); plus, only where the class sweep in the dispatch finds another refused summary × field-type pair, that other src/*/views/*.view.ts site, except any file an open PR holds; one .changeset/1980-*.md (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: default (dispatch-gates --tier --repo objectstack-ai/hotcrm answers "no path-derived mandate … the tier stays the PM's per-card judgment call"; the class sweep is a judgement surface, so not the floor)
Clause-②: no
Thread-read: none
Serial constraints cleared: in-flight #1984 (one-off, session_01WXp8E7s1gzSBje9ZwwqVux) declares src/sales/flows/forecast-snapshot.flow.ts, the same object but another file; disjoint. The open PRs' file lists (read via pulls/N/files: #1981 #1977 #1963 #1962 #1953 #1950 #1058) do not carry forecast.view.ts. #1950 (draft, Track A) holds src/sales/views/opportunity.view.ts, so the sweep reports and does not edit there. The sibling dispatch #1974 is disjoint.

First-touch grading. This card arrived with no labels. It now carries bug and priority:p3, with route drop the summary on the formula column.

• Provenance: the maintainer was asked in this seat's chat at 2026-10-02T21:21Z and chose, verbatim, 「p3 bug · 去掉该列合计(推荐)」.
• Why p3: today the summary is a client-side footer, so the refusal is invisible and no user sees a wrong number. The harm is that the exhibit teaches a pair the platform refuses.
• Why not a stored currency field: it adds a writer, and that writer sits on Move the six flow value expressions os validate --strict flags from the {…} template dialect to CEL envelopes, with tests #1984's in-flight forecast-snapshot.flow.ts.

repo:hotcrm seat · 2026-10-02T21:52Z

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report
{
"issue": 1980,
"status": "done",
"branch": "claude/issue-1980-forecast-formula-summary",
"pr": "#1986",
"session": "session_01ER8ntXZhYebyQ66aXWdjfT (subagent of the repo:hotcrm seat; container CLAUDE_CODE_REMOTE_SESSION_ID cse_01ER8ntXZhYebyQ66aXWdjfT)",
"premise_still_valid": true,
"summary": "Per the maintainer ruling (p3 bug, drop the column total), removed only the summary: 'sum' key on the expected_amount column of the grouped all_forecasts grid in src/sales/views/forecast.view.ts and added a one-line comment saying why; the column and the Quota/Closed Won/Commit/Best Case/Pipeline totals stay. Premise re-measured on origin/main cf422b8 (main moved past the dispatch base 39ba05e via #1985, which touches forecast-snapshot.flow.ts only): line 28 still carried the pair, forecast.object.ts:177 declares Field.formula, and the INSTALLED @objectstack/spec 17.6.0 ships AGGREGATE_FIELD_TYPE_COMPATIBILITY (dist/data/index.mjs) whose sum row is [number,currency,rating,slider,progress,summary,boolean,toggle], so formula (and percent) is refused. Class sweep over all 20 column summaries in the 14 registered view sets found exactly one refused pair (this one); no other file edited, and src/sales/views/opportunity.view.ts (#1950) has zero hits (both expected_revenue columns are currency). No test pins these summaries; docs promise no Expected total, so no doc change. Patch changeset .changeset/1980-forecast-formula-summary.md added.",
"tests": "HEAD 3cba8f5 (the sha pnpm verify ran on, read before the run and again after; worktree clean). OS_VERIFY_LOCK_SLOT=hotcrm-issue-1980 bash /home/user/objectstack/scripts/pm/os-verify-lock.sh -c 'pnpm verify' → os-verify-lock: VERDICT command-exit 0 · held the lock 169s (2m49s) · waited 0s. Stages: validate (18 Objects, 14 Views; 7 pre-existing author-time warnings, none from this diff) · typecheck clean · lint 1 warning(s), 16 suggestion(s) · ✓ i18n lint gate: 0 i18n/missing-* issues · ✓ source hygiene clean · ✓ source token ratchet clean · build Artifact: dist/objectstack.json · vitest Test Files 174 passed (174) / Tests 3728 passed | 1 skipped (3729). Class-sweep instrument (uncommitted tsx: imports allViews/allObjects from objectstack.composition.ts, maps member→function via COLUMN_SUMMARY_AGGREGATION copied from installed dist/ui/index.mjs, judges with the spec's isAggregateCompatibleWithFieldType): base cf422b8 = 20 sites, REFUSED pairs: 1 (positive control: crm_forecast:all_forecasts expected_amount sum→sum x formula REFUSED); after 3cba8f5 = 19 sites, REFUSED pairs: 0. Instrument coverage cross-check: git grep -n \"summary:\" -- 'src/*/views/*.view.ts' = 20 lines (18 sum, 2 avg), equal to the instrument's 20. No ablation: no test or guard was added (AGENTS.md scope rule 3), so the before/after sweep is the measurement. Control-byte self-scan on both changed files: no hits (grep exit 1). CI at report time (one read, not polled): Check Changeset success, Quality Checks success, Label Pull Request success, link-check success, Vercel Preview Comments success; Build and Test (22.x), Playwright, Analyze Code in_progress.",
"token_ratchet": "src/sales (only package touched): business semantics ~54,323 → ~54,323 (ceiling ~55,000); interaction layer ~28,003 → ~27,999 (ceiling ~31,000); authored total ~97,805 → ~97,801 (ceiling ~100,000). Base read on untouched cf422b8, after read from the pnpm verify run at 3cba8f5. No ceiling changed; service/revenue/marketing untouched.",
"files_changed": [
"src/sales/views/forecast.view.ts (+2 -1: summary key removed on expected_amount, one comment line added)",
".changeset/1980-forecast-formula-summary.md (new, 'hotcrm': patch)"
],
"deviations": [
"Commit 3cba8f5 carries a model-named trailer (Co-Authored-By: Claude Opus 5.5), copied from the harness attribution reminder; the dispatch contract asks for model-free trailer pairs. History not rewritten (no force-push grant in hotcrm AGENTS.md); the squash-merge message can drop that line.",
"Worked from origin/main cf422b8, not the dispatch base 39ba05e: #1985 landed in between (forecast-snapshot.flow.ts, quote-generation.flow.ts, two tests, a changeset), disjoint from this surface.",
"pnpm verify was launched under the lock with run_in_background plus its own log and pid file, then waited on in the foreground with tail --pid inside this turn (the sanctioned blocking-wait shape), to stay clear of the ~10-minute foreground cap; it finished in 2m49s."
],
"mcp_calls": "0",
"api_writes": "3 — each sent as ONE repository_dispatch to objectstack-ai/objectstack (HTTP 204) and executed by the fleet-write relay as objectstack-fleet[bot]: (1) pr_create → POST /repos/objectstack-ai/hotcrm/pulls (draft #1986; relay read-back 6844 bytes sent, 6844 stored, identical; own REST read-back identical too); (2) label-write --assign hotlong → POST /repos//issues/1986/assignees (read-back MATCHES: assignees hotlong, labels metadata unchanged); (3) post-stamped --comment=1980 → POST /repos//issues/1980/comments (this report). Not REST: two git pushes of the branch (empty probe, then 3cba8f5). Reads: GET issue 1980 + comments, GET pulls/1986, GET check-runs for 3cba8f5.",
"open_questions": [],
"out_of_scope_findings": [
"carrier: 承接者:无 · noted, not filed · scripts/check-source-token-ratchet.mjs prints, on both base and head, headroom is 2,997 tokens, over twice the 5% buffer — re-anchor this ceiling to ~30,000 for src/sales interaction layer; outside this card's claimed surface (forecast.view.ts + one changeset) and a sibling hotcrm dev shares the container, so left alone. In PR Acceptance notes only.",
"carrier: 承接者:无 · noted, not filed · NOT MEASURED: whether the console's grouped All Forecasts view actually issues the spec's compileListViewGroupQuery (17.6.0 dist/ui/index.mjs compiles a grouped view's column summary into a server-side sum_expected_amount node) — the card states the summary is a client-side footer today. Moot for this card: the node is gone with the fix. In PR Acceptance notes only.",
"opportunity.view.ts (#1950, Track A): sweep found 0 refused pairs there (amount and expected_revenue are currency on both open_opportunities and all_opportunities) — nothing to carry."
]
}

---

Generated by Claude Code

[objectstack-fleet]

ACCEPT: PR #1986 at 3cba8f5f. The expected_amount column keeps its values and loses only its sum total. repo:hotcrm seat, session_01ER8ntXZhYebyQ66aXWdjfT, 2026-10-02T22:07Z. Report: 5962208619.

• Form: draft, base main. The first line is Fixes #1980, and it is the body's only closing keyword. The PR assignee is hotlong, mirroring the card.

• Scope: 2 files. src/sales/views/forecast.view.ts (+2/−1: the key is removed, and a one-line comment says why) and .changeset/1980-forecast-formula-summary.md ('hotcrm': patch). Nothing else.

• Changeset, sentence by sentence:

  • The Expected field is Field.formula over closed_amount + commit_amount (forecast.object.ts:177).
  • The refusal is real: the installed @objectstack/spec@17.6.0 dist/data/index.mjs:21217 reads sum: [...ADDITIVE_AGGREGATE_FIELD_TYPES, ...BOOLEAN_AGGREGATE_FIELD_TYPES], and formula is in neither list.
  • Closed Won, Commit, Quota, Best Case and Pipeline keep summary: 'sum' in the diff. All accurate.

• Class sweep, re-counted by the seat: git grep 'summary:' origin/main -- 'src/*/views/*.view.ts' = 20 lines. That equals the 20 sites the dev's instrument judged, with the known pair as the positive control. 1 refused pair before, 0 after. opportunity.view.ts (held by feat(opportunity): REQ-0006 qualification, the customer calendar, narrative and approval on status change #1950) has 0 refused pairs and is untouched.

• Gates: the dev's pnpm verify at 3cba8f5f ended with VERDICT command-exit 0: 174 test files, 3,728 tests. Token ratchet: src/sales interaction layer −4, other layers unchanged, no ceiling moved. CI at 2026-10-02T22:06Z: 8 of 9 success on 3cba8f5f, and Build and Test (22.x) was still in_progress. ⛔ Landing waits for it.

• Deviation, recorded and not REWORK: commit 3cba8f5f carries a model-named Co-Authored-By trailer, copied from the harness's attribution reminder. The dev reported it and did not rewrite history, as its rules require.

• Out of scope:

  • A ratchet re-anchor advisory on src/sales (interaction layer headroom 2,997). Not a defect.
  • Whether the console compiles a grouped column summary server-side: NOT MEASURED, and moot after this fix.

  Both are noted, not filed: neither is in the three filing classes.

Landing: no governed path, so once every check is green the seat readies the PR and arms auto-merge, and the merge queue lands it. MERGED then closes this card.

---

Generated by Claude Code