feat: calibration from closed-issue actuals → forecast flips off cold-start (#1)

Close the D3 loop. The forecast now learns from the team's own estimate-vs-actual
history (the working time #5 infers from git events) instead of guessing forever.

core (@commitea/core/calibration-v0):
- fitCalibration(samples): lognormal fit on log(actual/estimate) — global +
  per-bucket (once a bucket clears the floor) + per-person bias. coldStart until
  n >= 20 closed-with-estimate issues.
- calibrationSamples(): pull those samples from the closed backlog via lifecycle
  inference (estimate label vs inferred actualWorkingDays).
- toDurationModel(): project the fit to the params forecast consumes.
- forecast() gains options.model: when past cold-start, fitted params drive the
  sim (per bucket, global fallback); otherwise the code priors do. Forecast.coldStart
  now reflects the model. nearestBucket extracted + exported.

app:
- AppShell fits calibration once from the reconciled backlog, feeds the model into
  forecastBacklog (cone), and drives the Calibration screen + Runway header.
- Focus cone footer, Runway note, and Calibration screen now say cold-start (N/20)
  vs calibrated (on N closed) from real data; Calibration scatter / bucket bias /
  per-person all fitted, degrading honestly on a thin dataset.

Known refinement: same-day closes yield 0 working-day actuals (day-granular) and
are excluded, so a fast-moving repo can sit at n=0 — honest, but a fractional
(hours-based) actual would let those count. Per-person uses gitea login, not
display name, until the person map lands.

Note: also re-lands #10 (Monte Carlo) and #5 (lifecycle) which merged into their
stacked base branches but never propagated to main (stacked-merge trap); this
branch is cut from main and carries all three so main is whole again.

Verified: 74 core tests green (9 calibration + 2 forecast-switch added), desktop
typecheck clean, 14 fixture e2e green, live spec asserts the real cold-start
calibration surface (Runway note + screen badge fitted from actuals).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

packages/core/src/calibration/calibration-v0.test.ts

@@ -0,0 +1,121 @@
+import { describe, expect, it } from 'vitest'
+
+import { extractLabelFacts } from '../labels/label-schema.js'
+import type { LifecycleEvent } from '../lifecycle/lifecycle-v0.js'
+import type { GiteaIssue } from '../gitea/types.js'
+import {
+  CALIBRATION_BUCKET_FLOOR,
+  calibrationSamples,
+  type CalibrationSample,
+  COLD_START_THRESHOLD,
+  fitCalibration,
+  toDurationModel,
+} from './calibration-v0.js'
+
+function sample(over: Partial<CalibrationSample> = {}): CalibrationSample {
+  return { issue: 1, estimateDays: 2, actualWorkingDays: 2, bucket: 2, person: null, ...over }
+}
+
+describe('fitCalibration', () => {
+  it('is cold-start below the threshold and reports the honest n', () => {
+    const m = fitCalibration([sample(), sample({ actualWorkingDays: 4 })])
+    expect(m.n).toBe(2)
+    expect(m.coldStart).toBe(true)
+  })
+
+  it('flips off cold-start at the threshold', () => {
+    const many = Array.from({ length: COLD_START_THRESHOLD }, (_, i) =>
+      sample({ issue: i, estimateDays: 2, actualWorkingDays: 3, bucket: 2 }),
+    )
+    const m = fitCalibration(many)
+    expect(m.n).toBe(COLD_START_THRESHOLD)
+    expect(m.coldStart).toBe(false)
+  })
+
+  it('recovers the global median ratio (mu = mean log-ratio)', () => {
+    // every actual is exactly 2x its estimate → mu = ln 2
+    const m = fitCalibration(Array.from({ length: 25 }, (_, i) => sample({ issue: i, estimateDays: 2, actualWorkingDays: 4 })))
+    expect(m.global.mu).toBeCloseTo(Math.log(2), 6)
+  })
+
+  it('fits a bucket only once it clears the floor', () => {
+    const twos = Array.from({ length: CALIBRATION_BUCKET_FLOOR }, (_, i) =>
+      sample({ issue: i, estimateDays: 2, actualWorkingDays: 3, bucket: 2 }),
+    )
+    const oneThin = [sample({ issue: 99, estimateDays: 5, actualWorkingDays: 9, bucket: 5 })]
+    const m = fitCalibration([...twos, ...oneThin])
+    expect(m.byBucket[2]?.n).toBe(CALIBRATION_BUCKET_FLOOR)
+    expect(m.byBucket[5]).toBeUndefined() // only 1 sample, below floor
+  })
+
+  it('drops non-positive estimates/actuals', () => {
+    const m = fitCalibration([sample({ actualWorkingDays: 0 }), sample({ estimateDays: 0 }), sample()])
+    expect(m.n).toBe(1)
+  })
+
+  it('derives a per-person bias relative to global', () => {
+    // one person consistently runs longer than the mean
+    const base = Array.from({ length: 20 }, (_, i) => sample({ issue: i, actualWorkingDays: 2, person: 'ak' }))
+    const slow = Array.from({ length: 3 }, (_, i) => sample({ issue: 100 + i, actualWorkingDays: 6, person: 'sm' }))
+    const m = fitCalibration([...base, ...slow])
+    expect(m.byPerson['sm'].biasMu).toBeGreaterThan(0)
+    expect(m.byPerson['ak'].biasMu).toBeLessThan(0)
+  })
+})
+
+describe('toDurationModel', () => {
+  it('projects the fit down to the params forecast needs', () => {
+    const m = fitCalibration(
+      Array.from({ length: 25 }, (_, i) => sample({ issue: i, estimateDays: 2, actualWorkingDays: 3, bucket: 2 })),
+    )
+    const dm = toDurationModel(m)
+    expect(dm.coldStart).toBe(false)
+    expect(dm.byBucket[2].mu).toBeCloseTo(m.byBucket[2].mu, 6)
+    expect(dm.global.mu).toBeCloseTo(m.global.mu, 6)
+  })
+})
+
+describe('calibrationSamples', () => {
+  const asOf = new Date('2026-02-01T00:00:00Z')
+
+  function issue(over: Partial<GiteaIssue>): GiteaIssue {
+    const labels = over.labels ?? []
+    return {
+      number: 1,
+      title: '#1',
+      body: '',
+      state: 'closed',
+      labels,
+      facts: extractLabelFacts(labels),
+      milestone: null,
+      assignee: null,
+      assignees: [],
+      createdAt: '2026-01-05T09:00:00Z',
+      updatedAt: '2026-01-12T09:00:00Z',
+      closedAt: '2026-01-12T09:00:00Z',
+      url: '',
+      ...over,
+    }
+  }
+
+  const events = (i: number): Record<number, LifecycleEvent[]> => ({
+    [i]: [{ type: 'commit', at: '2026-01-07T09:00:00Z' }, { type: 'close', at: '2026-01-12T09:00:00Z' }],
+  })
+
+  it('samples closed, estimated issues with a resolvable actual', () => {
+    const i = issue({ number: 7, labels: ['est/2d'], assignee: 'sm', closedAt: '2026-01-12T09:00:00Z' })
+    const [s] = calibrationSamples([i], events(7), asOf)
+    expect(s.issue).toBe(7)
+    expect(s.estimateDays).toBe(2)
+    expect(s.bucket).toBe(2)
+    expect(s.person).toBe('sm')
+    // Wed 2026-01-07 → Mon 2026-01-12 = Wed,Thu,Fri = 3 working days
+    expect(s.actualWorkingDays).toBe(3)
+  })
+
+  it('skips open issues and closed ones without an estimate', () => {
+    const open = issue({ number: 8, state: 'open', labels: ['est/2d'], closedAt: null })
+    const noEst = issue({ number: 9, labels: [] })
+    expect(calibrationSamples([open, noEst], { ...events(8), ...events(9) }, asOf)).toEqual([])
+  })
+})