fix(plan): move the Copilot credits bar onto the estimate and print whole credits (#1199)

docs/providers/copilot.md

@@ -160,7 +160,8 @@ see the #927 ruling in `src/session-cache.ts`).
   figure)`), and `codeburn status --format json | jq .plans.copilot` carries
   `spentCredits` (exact only), `estimatedCredits` (exact plus estimate),
   `creditRatedCalls` / `creditUnratedCalls`, `creditsIncomplete` and a plain
-  `creditsNote`. The bar and `percentUsed` deliberately stay on the exact figure.
+  `creditsNote`. The bar and `percentUsed` follow `estimatedCredits` while any
+  request is unrated, and the exact figure once every request carries one.
   For the live authoritative number, use the menubar's GitHub quota endpoint
   (separate system, PR #1200).
 - **Sync.** `codeburn sync push` holds a copilot session until it has been

src/dashboard.tsx

@@ -557,9 +557,9 @@ export function planBudgetHeadline(planUsage: PlanUsage): string {
     if (unrated > 0) {
       const rated = planUsage.creditRatedCalls ?? 0
       const estimated = planUsage.estimatedCredits ?? spent
-      // Tenths, not the exact-figure precision: the bulk of this number is
-      // priced from tokens, so more digits would claim accuracy it lacks.
-      const shown = formatCredits(Math.round(estimated * 10) / 10)
+      // Whole credits only: the bulk of this number is priced from tokens,
+      // so decimals would claim accuracy it lacks.
+      const shown = formatCredits(Math.round(estimated))
       return `${planLabel(planUsage)}: ~${shown} / ${formatCredits(budget)} AI Credits (estimated; ${rated} of ${rated + unrated} requests carry GitHub's exact figure)`
     }
     return `${planLabel(planUsage)}: ${formatCredits(spent)} / ${formatCredits(budget)} AI Credits`

src/plan-usage.ts

@@ -152,7 +152,7 @@ export function copilotCreditsNote(rated: number, unrated: number): string {
   if (unrated === 0) {
     return `All ${rated} Copilot requests in this period carry GitHub's exact credit figure, so spentCredits is complete.`
   }
-  return `${rated} of ${rated + unrated} Copilot requests carry GitHub's exact credit figure (Copilot CLI session-store rows are the only local source that has it). spentCredits counts only those; estimatedCredits adds the rest priced from tokens at listed API rates, which is a floor because request multipliers and cached-token pricing are not modelled. percentUsed still tracks spentCredits.`
+  return `${rated} of ${rated + unrated} Copilot requests carry GitHub's exact credit figure (Copilot CLI session-store rows are the only local source that has it). spentCredits counts only those; estimatedCredits adds the rest priced from tokens at listed API rates, which is a floor because request multipliers and cached-token pricing are not modelled. percentUsed tracks estimatedCredits while any request is unrated.`
 }
 
 export function projectMonthEnd(
@@ -203,7 +203,11 @@ export function getPlanUsageFromProjects(plan: Plan, projects: ProjectSummary[],
   if (isCopilotCreditsPlan(plan)) {
     const { spentCredits, estimatedCredits, creditRatedCalls, creditUnratedCalls, creditsIncomplete } = copilotCreditSpend(projects)
     const budgetCredits = plan.monthlyCredits ?? 0
-    const percentUsed = budgetCredits > 0 ? (spentCredits / budgetCredits) * 100 : 0
+    // The bar follows the estimate once any request lacks an exact figure:
+    // a bar at 0% beside a 19k real bill is the misleading picture #1199
+    // reported, and the estimate is a floor, so it under-alerts, never over.
+    const barCredits = creditUnratedCalls > 0 ? estimatedCredits : spentCredits
+    const percentUsed = budgetCredits > 0 ? (barCredits / budgetCredits) * 100 : 0
     const spent = creditsToUsd(spentCredits)
     const status: PlanStatus = percentUsed > 100 ? 'over' : percentUsed >= PLAN_NEAR_THRESHOLD_PCT ? 'near' : 'under'
     const projectedMonthUsd = projectMonthEnd(projects, periodStart, periodEnd, today, spent, call => planCallSpend(plan, call))

tests/plan-budget-copy.test.ts

@@ -102,7 +102,7 @@ describe('plan budget copy', () => {
       creditUnratedCalls: 469,
     })
     expect(planBudgetHeadline(copilot))
-      .toBe("Copilot Max: ~9800.4 / 20000 AI Credits (estimated; 4 of 473 requests carry GitHub's exact figure)")
+      .toBe("Copilot Max: ~9800 / 20000 AI Credits (estimated; 4 of 473 requests carry GitHub's exact figure)")
     expect(planBudgetHeadline(copilot)).not.toContain('—')
   })
 

tests/plan-usage.test.ts

@@ -504,7 +504,7 @@ describe('copilot AI credit plan math', () => {
     expect(usage.status).toBe('under')
   })
 
-  it('does not fill the credits bar from token costUSD when nanoAiu is missing', () => {
+  it('fills the credits bar from the token estimate when nanoAiu is missing', () => {
     const usage = getPlanUsageFromProjects(copilotPro, [
       usageProject([
         usageCall({
@@ -517,7 +517,9 @@ describe('copilot AI credit plan math', () => {
 
     expect(usage.spentCredits).toBe(0)
     expect(usage.creditsIncomplete).toBe(true)
-    expect(usage.percentUsed).toBe(0)
+    // 42 USD at 0.01 USD per credit against the 1500-credit Pro budget.
+    expect(usage.percentUsed).toBe(280)
+    expect(usage.status).toBe('over')
     expect(usage.spentApiEquivalentUsd).toBe(0)
   })
 
@@ -618,7 +620,7 @@ describe('copilot AI credit plan math', () => {
     expect(spend.creditsIncomplete).toBe(true)
   })
 
-  it('leaves the bar and percentUsed on the exact figure while exposing the estimate', () => {
+  it('moves the bar and percentUsed onto the estimate while exposing the exact figure', () => {
     const usage = getPlanUsageFromProjects(copilotPro, [
       usageProject([
         usageCall({ provider: 'copilot', costUSD: 0.42, timestamp: '2026-08-05T12:00:00.000Z' }),
@@ -626,7 +628,7 @@ describe('copilot AI credit plan math', () => {
     ], today)
 
     expect(usage.spentCredits).toBe(0)
-    expect(usage.percentUsed).toBe(0)
+    expect(usage.percentUsed).toBeCloseTo(2.8, 10)
     expect(usage.estimatedCredits).toBeCloseTo(42, 10)
     expect(usage.creditUnratedCalls).toBe(1)
   })