Merge pull request #560 from objectstack-ai/claude/hooks-flows-automation-conflict-isglbj

fix(opportunity): derive days_in_stage from a stored stage_entry_date

Author: yinlianghui

.changeset/opportunity-stage-entry-date.md

@@ -0,0 +1,24 @@
+---
+'hotcrm': patch
+---
+
+Make the stalled-deal sweep actually fire (#489). `crm_opportunity.days_in_stage`
+was a plain number column that nothing ever incremented — the lifecycle hook
+reset it to 0 on a stage change and no sweep raised it — so
+`opportunity_stagnation`'s `days_in_stage > 14` filter matched only the rows the
+seed had hardcoded, and real deals could rot in a stage forever without a nudge.
+
+Opportunities now carry `stage_entry_date`, a stored date stamped by the
+lifecycle hook on insert and on every stage change. `days_in_stage` becomes a
+formula counting from it, correct on every read with no nightly full-table pass.
+Because a formula is evaluated after the query and is not a real column, the
+sweep now predicates on `stage_entry_date < TODAY() − 14` — the same test,
+against something the data engine can see — and the "Stale Opportunities" view,
+which can express neither a formula predicate nor a relative date, becomes an
+open-deals list ordered longest-in-stage first with `days_in_stage` on show.
+
+Two guards in `test/metadata-references.test.ts` keep the class from returning:
+no list view may filter or sort on a formula field, and no flow data node may
+filter on one. The second instance they caught is fixed here too — the quarterly
+forecast view's `attainment_pct` tiebreaker was a dead sort key, now
+`closed_amount`.

content/docs/sales/opportunities.mdx

@@ -31,7 +31,7 @@ The detail screen has 7 sections:
 | --- | --- |
 | **Basic Information** | Name, account, primary contact, owner |
 | **Financials** | Amount, expected revenue *(auto)*, probability *(auto from stage)* |
-| **Sales Process** | Stage, close date, created date, days in current stage |
+| **Sales Process** | Stage, close date, created date, stage entry date |
 | **Classification** | Type (New Business / Upgrade / Renewal / Expansion), lead source |
 | **Competition & Campaigns** | Competitors, source campaign |
 | **Notes & Next Steps** | Description, next steps |
@@ -54,7 +54,7 @@ You don't have to remember any of this — the system does it for you:
 
 - **Probability** is set from the stage whenever you move the deal (you *can* override it manually if you have a special situation, but it'll reset next time the stage changes).
 - **Expected revenue** = amount × probability ÷ 100, recalculated whenever amount or probability changes.
-- **Days in current stage** is recomputed daily — it surfaces stagnant deals.
+- **Stage entry date** is stamped the moment a deal is created and re-stamped every time you move it to a new stage. **Days in current stage** counts from it, so it is always up to date — that is what surfaces stagnant deals.
 - When you mark a deal **Closed Won and the amount is over $100K**, sales management gets an email.
 - When the amount exceeds **$500K**, the deal enters the [Discount Approval](/docs/revenue/approvals) process automatically.
 

content/docs/sales/opportunities.zh-Hans.mdx

@@ -31,7 +31,7 @@ description: 活跃的销售交易——销售管道的核心，包含 7 个阶
 | --- | --- |
 | **基本信息** | 名称、客户、主要联系人、负责人 |
 | **财务** | 金额、预期营收 *(自动)*、概率 *(从阶段自动得出)* |
-| **销售流程** | 阶段、成交日期、创建日期、当前阶段停留天数 |
+| **销售流程** | 阶段、成交日期、创建日期、进入当前阶段日期 |
 | **分类** | 类型（新业务 / 升级 / 续约 / 扩展）、线索来源 |
 | **竞争与营销活动** | 竞争对手、来源营销活动 |
 | **备注与后续步骤** | 描述、后续步骤 |
@@ -54,7 +54,7 @@ description: 活跃的销售交易——销售管道的核心，包含 7 个阶
 
 - 每当你推进交易时，**概率**会从阶段中设定（如果你有特殊情况，*可以*手动覆盖，但下次阶段变化时它会重置）。
 - **预期营收** = 金额 × 概率 ÷ 100，每当金额或概率变化时重新计算。
-- **当前阶段停留天数**每天重新计算——它能暴露停滞的交易。
+- **进入当前阶段日期**在交易创建时打上，之后每次移动到新阶段都会重新打上。**当前阶段停留天数**由它推算，因此始终是最新的——这正是暴露停滞交易的依据。
 - 当你将一笔交易标记为**赢单且金额超过 $100K** 时，销售管理层会收到邮件。
 - 当金额超过 **$500K** 时，交易会自动进入 [折扣审批](/zh-Hans/docs/revenue/approvals) 流程。
 

content/docs/sales/opportunities.zh-Hant.mdx

@@ -31,7 +31,7 @@ description: 活躍的銷售交易——銷售管道的核心，包含 7 個階
 | --- | --- |
 | **基本資訊** | 名稱、客戶、主要聯絡人、負責人 |
 | **財務** | 金額、預期營收 *(自動)*、機率 *(從階段自動得出)* |
-| **銷售流程** | 階段、成交日期、建立日期、當前階段停留天數 |
+| **銷售流程** | 階段、成交日期、建立日期、進入當前階段日期 |
 | **分類** | 類型（新業務 / 升級 / 續約 / 擴展）、潛在客戶來源 |
 | **競爭與行銷活動** | 競爭對手、來源行銷活動 |
 | **備註與後續步驟** | 描述、後續步驟 |
@@ -54,7 +54,7 @@ description: 活躍的銷售交易——銷售管道的核心，包含 7 個階
 
 - 每當你推進交易時，**機率**會從階段中設定（如果你有特殊情況，*可以*手動覆蓋，但下次階段變化時它會重置）。
 - **預期營收** = 金額 × 機率 ÷ 100，每當金額或機率變化時重新計算。
-- **當前階段停留天數**每天重新計算——它能暴露停滯的交易。
+- **進入當前階段日期**在交易建立時打上，之後每次移動到新階段都會重新打上。**當前階段停留天數**由它推算，因此始終是最新的——這正是暴露停滯交易的依據。
 - 當你將一筆交易標記為**贏單且金額超過 $100K** 時，銷售管理層會收到郵件。
 - 當金額超過 **$500K** 時，交易會自動進入 [折扣審批](/zh-Hant/docs/revenue/approvals) 流程。
 

docs/developers/api_reference.md

@@ -36,7 +36,7 @@ Source: `src/objects/opportunity.object.ts`
 
 Key fields:
 
-`name`, `crm_account`, `primary_contact`, `owner`, `amount`, `expected_revenue`, `stage`, `probability`, `close_date`, `created_date`, `type`, `lead_source`, `competitors`, `crm_campaign`, `days_in_stage`, `next_step`, `is_private`, `forecast_category`, `approval_status`, `approved_date`, `win_reason`, `loss_reason`, `loss_details`
+`name`, `crm_account`, `primary_contact`, `owner`, `amount`, `expected_revenue`, `stage`, `probability`, `close_date`, `created_date`, `stage_entry_date`, `type`, `lead_source`, `competitors`, `crm_campaign`, `days_in_stage`, `next_step`, `is_private`, `forecast_category`, `approval_status`, `approved_date`, `win_reason`, `loss_reason`, `loss_details`
 
 ### `crm_opportunity_line_item` - Opportunity Line Item
 

src/data/index.ts

@@ -38,7 +38,7 @@ const celDaysFromNow = (n: number) => cel`daysFromNow(${n})`;
  * Two consequences shape everything below:
  *
  * 1. Seeding historical values into readonly fields (`created_date`,
- *    `days_in_stage`, `last_contacted_date`, `actual_revenue`) is legitimate
+ *    `stage_entry_date`, `last_contacted_date`, `actual_revenue`) is legitimate
  *    and load-bearing — it is the only way demo reports get history — and the
  *    platform explicitly preserves explicit seed values.
  * 2. BUT nothing recomputes derived fields for seed rows, so every seeded
@@ -325,6 +325,14 @@ const leads = defineSeed(Lead, {
 // don't run over seeds, so every row below carries the exact values the hook
 // would compute — anything else gets silently rewritten on the first user
 // edit (#490).
+//
+// `stage_entry_date` follows the same rule: the hook stamps it on every insert
+// and stage change, so a seeded deal without one is a shape real data never
+// has — and it is what the stagnation sweep filters on and what the
+// `days_in_stage` FORMULA counts from, so it replaces the old hardcoded
+// `days_in_stage` numbers (#489). Open deals carry an age that spans the
+// 14-day stale threshold in both directions; settled deals entered their
+// closed stage on their close date.
 const opportunities = defineSeed(Opportunity, {
   mode: 'upsert',
   externalId: 'name',
@@ -340,7 +348,7 @@ const opportunities = defineSeed(Opportunity, {
       type: 'existing_upgrade',
       forecast_category: 'commit',
       lead_source: 'web',
-      days_in_stage: 12,
+      stage_entry_date: celDaysAgo(12),
       description: `Upgrade from Standard to Enterprise edition for the
 NA + EMEA teams. Drivers: (1) AI agent governance becomes a hard
 requirement after their internal compliance review, (2) advanced
@@ -358,7 +366,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       type: 'new_business',
       forecast_category: 'pipeline',
       lead_source: 'referral',
-      days_in_stage: 45,
+      stage_entry_date: celDaysAgo(45),
     },
     {
       name: 'Wayne Enterprise License',
@@ -371,7 +379,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       type: 'new_business',
       forecast_category: 'commit',
       lead_source: 'partner',
-      days_in_stage: 7,
+      stage_entry_date: celDaysAgo(7),
     },
     {
       name: 'Initech Cloud Migration',
@@ -384,7 +392,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       type: 'existing_upgrade',
       forecast_category: 'best_case',
       lead_source: 'event',
-      days_in_stage: 38,
+      stage_entry_date: celDaysAgo(38),
     },
     // ─── Closed Won deals (powers KPIs & revenue trends) ────────────────
     {
@@ -395,6 +403,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       probability: 100,
       expected_revenue: 220000,
       close_date: cel`daysAgo(15)`,
+      stage_entry_date: cel`daysAgo(15)`,
       type: 'existing_renewal',
       forecast_category: 'closed',
       lead_source: 'partner',
@@ -408,6 +417,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       probability: 100,
       expected_revenue: 145000,
       close_date: cel`daysAgo(50)`,
+      stage_entry_date: cel`daysAgo(50)`,
       type: 'new_business',
       forecast_category: 'closed',
       lead_source: 'event',
@@ -420,6 +430,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       probability: 100,
       expected_revenue: 380000,
       close_date: cel`daysAgo(95)`,
+      stage_entry_date: cel`daysAgo(95)`,
       type: 'existing_upgrade',
       forecast_category: 'closed',
       lead_source: 'web',
@@ -432,6 +443,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       probability: 100,
       expected_revenue: 65000,
       close_date: cel`daysAgo(140)`,
+      stage_entry_date: cel`daysAgo(140)`,
       type: 'new_business',
       forecast_category: 'closed',
       lead_source: 'referral',
@@ -444,6 +456,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       probability: 100,
       expected_revenue: 90000,
       close_date: cel`daysAgo(200)`,
+      stage_entry_date: cel`daysAgo(200)`,
       type: 'new_business',
       forecast_category: 'closed',
       lead_source: 'web',
@@ -457,6 +470,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       probability: 0,
       expected_revenue: 0,
       close_date: cel`daysAgo(25)`,
+      stage_entry_date: cel`daysAgo(25)`,
       type: 'existing_upgrade',
       forecast_category: 'omitted',
       lead_source: 'cold_call',
@@ -470,6 +484,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
       probability: 0,
       expected_revenue: 0,
       close_date: cel`daysAgo(60)`,
+      stage_entry_date: cel`daysAgo(60)`,
       type: 'new_business',
       forecast_category: 'omitted',
       lead_source: 'advertisement',
@@ -517,6 +532,12 @@ analytics seats for the Ops org, (3) priority support SLA.`,
         const close_date = isClosed(stage)
           ? celDaysAgo(5 + spread)
           : celDaysFromNow(7 + spread);
+        // A settled deal entered its closed stage the day it closed; an open
+        // one has been parked 3–62 days, straddling STALE_THRESHOLD_DAYS so
+        // the stagnation sweep and the stale view both have rows on each side.
+        const stage_entry_date = isClosed(stage)
+          ? celDaysAgo(5 + spread)
+          : celDaysAgo(3 + (i * 11) % 60);
         const amount = 20000 + ((i * 17_393) % 480_000);
         const probability = probabilityByStage[stage];
         out.push({
@@ -531,7 +552,7 @@ analytics seats for the Ops org, (3) priority support SLA.`,
           type: types[i % types.length],
           forecast_category: forecastByStage[stage],
           lead_source: sources[i % sources.length],
-          ...(isClosed(stage) ? {} : { days_in_stage: 3 + (i * 11) % 60 }),
+          stage_entry_date,
         });
       }
       return out;

src/flows/opportunity-stagnation.flow.ts

@@ -6,8 +6,9 @@ type Flow = Automation.Flow;
 /**
  * Opportunity Stagnation — scheduled "deal-rot" detector.
  *
- * The opportunity carries `days_in_stage`, but nothing acted on it, so deals
- * could sit untouched in a stage indefinitely. This daily sweep finds open
+ * The opportunity carries `stage_entry_date` (stamped by the lifecycle hook on
+ * insert and on every stage change), but nothing acted on it, so deals could
+ * sit untouched in a stage indefinitely. This daily sweep finds open
  * opportunities stalled longer than the threshold, nudges the owner and books
  * a follow-up task so the deal re-enters the working set. A deal with an
  * open stall task is skipped (idempotency), so each stall episode produces
@@ -38,9 +39,16 @@ export const OpportunityStagnationFlow: Flow = {
       id: 'query_stalled', type: 'get_record', label: 'Find Stalled Deals',
       config: {
         objectName: 'crm_opportunity',
+        // Predicate on the STORED `stage_entry_date`, not on `days_in_stage`:
+        // the latter is a formula, computed after the query, so as a filter key
+        // it addressed a column that does not exist (#489). `entry < today − N`
+        // is the same test as `days_in_stage > N`, resolved by the flow
+        // template engine (same `{TODAY() ± n}` token as contract-renewal).
+        // A row with a null `stage_entry_date` does not satisfy `$lt` and is
+        // skipped, preserving "nothing has stagnated yet" for unstamped rows.
         filter: {
           stage: { $nin: ['closed_won', 'closed_lost'] },
-          days_in_stage: { $gt: STALE_THRESHOLD_DAYS },
+          stage_entry_date: { $lt: `{TODAY() - ${STALE_THRESHOLD_DAYS}}` },
         },
         limit: 500,
         outputVariable: 'oppList',

src/objects/opportunity.hook.ts

@@ -7,6 +7,8 @@ import type { HookApi } from './_hook-api';
  * Opportunity lifecycle hook.
  *
  * - Re-derives `expected_revenue` from `amount * stageProbability` when either changes.
+ * - Stamps `stage_entry_date` on insert and on every stage change — the stored
+ *   clock behind the `days_in_stage` formula and the stagnation sweep (#489).
  * - Freezes most fields after stage is closed (won/lost) — only narrative fields editable.
  * - On `closed_won`: stamps `close_date=today`, promotes the parent account to `customer`,
  *   and asynchronously schedules an "Activate customer" task.
@@ -115,16 +117,27 @@ const opportunityValidationHook: Hook = {
       if (stageChanged) input.forecast_category = STAGE_FORECAST[stage];
     }
 
+    // Start the stage-age clock at creation. Without this the row lands with a
+    // null `stage_entry_date`, `days_in_stage` reads null instead of 0, and the
+    // deal is invisible to the stagnation sweep until its first stage change —
+    // exactly backwards, since a deal that never moves is the stalled one.
+    if (event === 'beforeInsert' && !input.stage_entry_date) {
+      input.stage_entry_date = new Date().toISOString().slice(0, 10);
+    }
+
     if (event === 'beforeUpdate' && previous) {
       // Stamp close_date when transitioning into closed_won
       if (input.stage === 'closed_won' && previous.stage !== 'closed_won' && !input.close_date) {
         input.close_date = new Date().toISOString().slice(0, 10);
       }
-      // Reset the stage-age counter on any stage change so a deal that
-      // advances stops matching the stagnation sweep. (Readonly fields are
-      // writable from before-hooks via input mutation.)
+      // Restart the stage-age clock on any stage change so a deal that
+      // advances stops matching the stagnation sweep. `days_in_stage` is a
+      // formula over this column, so this one write IS the reset — it used to
+      // set `days_in_stage = 0` against a counter nothing ever incremented.
+      // (Readonly fields are writable from before-hooks via input mutation:
+      // readonly stripping only drops keys the CALLER supplied.)
       if (typeof input.stage === 'string' && input.stage !== previous.stage) {
-        input.days_in_stage = 0;
+        input.stage_entry_date = new Date().toISOString().slice(0, 10);
       }
     }
   },

src/objects/opportunity.object.ts

@@ -1,4 +1,4 @@
-import { P, cel } from '@objectstack/spec';
+import { F, P, cel } from '@objectstack/spec';
 // Copyright (c) 2025 ObjectStack. Licensed under the Apache-2.0 license.
 
 import { ObjectSchema, Field } from '@objectstack/spec/data';
@@ -126,6 +126,20 @@ export const Opportunity = ObjectSchema.create({
       group: 'sales_process',
     }),
 
+    // Stage-age clock (#489). This is the STORED half of the pair: a real,
+    // indexed date column, so it is what automation and views may filter and
+    // sort on. `days_in_stage` below is a formula derived from it.
+    // `opportunity_lifecycle` stamps it on insert and re-stamps it on every
+    // stage change; readonly because nothing but that hook should move it
+    // (before-hook writes survive readonly stripping — only caller-supplied
+    // keys are dropped).
+    stage_entry_date: Field.date({
+      label: 'Stage Entry Date',
+      description: 'Date this opportunity entered its current stage.',
+      readonly: true,
+      group: 'sales_process',
+    }),
+
     // Additional Classification
     type: Field.select({
       label: 'Opportunity Type',
@@ -167,9 +181,27 @@ export const Opportunity = ObjectSchema.create({
     }),
 
     // Sales cycle metrics
-    days_in_stage: Field.number({
+    //
+    // FORMULA, not a stored counter (#489). As a plain number column nothing
+    // ever raised it: the hook reset it to 0 on a stage change and no sweep or
+    // hook anywhere incremented it, so `days_in_stage > 14` matched only the
+    // rows the seed hardcoded — the `opportunity_stagnation` flow never fired
+    // on real data. Deriving it from `stage_entry_date` is correct on every
+    // read and costs no nightly full-table pass.
+    //
+    // The trade-off: formulas are evaluated AFTER the query (the engine's
+    // `applyFormulaPlan` walks the returned rows), so this is not a real
+    // column and CANNOT appear in a filter or a sort. Anything that needs to
+    // *select* stalled deals predicates on `stage_entry_date` instead — see
+    // `opportunity-stagnation.flow.ts` and the `stale_opportunities` view.
+    //
+    // `has()` + null guard: `daysBetween(null, …)` faults and the whole field
+    // silently evaluates to null, so an unstamped row is spelled out as null
+    // rather than arriving there by accident.
+    days_in_stage: Field.formula({
       label: 'Days in Current Stage',
-      readonly: true,
+      expression: F`has(record.stage_entry_date) && record.stage_entry_date != null ? daysBetween(record.stage_entry_date, today()) : null`,
+      returnType: 'number',
       group: 'crm_forecast',
     }),
 
@@ -267,6 +299,8 @@ export const Opportunity = ObjectSchema.create({
     { fields: ['owner'] },
     { fields: ['stage'] },
     { fields: ['close_date'] },
+    // The stagnation sweep filters on this every morning.
+    { fields: ['stage_entry_date'] },
   ],
 
   // Enable advanced features

src/translations/en.ts

@@ -323,6 +323,7 @@ export const en: TranslationData = {
         competitors: { label: 'Competitors' },
         crm_campaign: { label: 'Campaign' },
         days_in_stage: { label: 'Days in Current Stage' },
+        stage_entry_date: { label: 'Stage Entry Date' },
         is_private: { label: 'Private' },
         approval_status: {
           label: 'Approval Status',