Compare commits

...

5 Commits

Author SHA1 Message Date
Croissant Le Doux
19e83ff8ba query_project: implement the standup view (#28)
The StandupScreen already renders drift + plan + nag from real data (standup-view,
#52), but the agent's query_project standup view was a `notImplemented` stub, so
Reginald couldn't answer standup questions from deterministic data.

Implement `standupView(snap, asOf)`: today's plan (the scheduler's earliest pick
per person, with why — critical path / blocks / order), overnight drift (real
anomalies: issues sitting in review, or steeping past their estimate), and the
single stalest blocker to nag about (+ what it blocks). All deterministic; the
model narrates. Added 'standup' to the query_project tool's view enum.

Acceptance met: standup surfaces schedule drift + at least one stale blocker
(and stays calm — no nag, empty drift — when nothing is steeping). +2 core tests;
typecheck clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-09 15:33:20 -04:00
2a6413821a Merge pull request 'calibration: count same-day closes honestly (#34)' (#54) from feat/calibration-honesty into main
Reviewed-on: #54
2026-07-09 19:23:59 +00:00
008435f1c2 Merge branch 'main' into feat/calibration-honesty 2026-07-09 19:23:55 +00:00
354ba9227e Merge pull request 'apply_changes: unified mutation tool — estimate/priority/assign/milestone, in-app + agent (#24)' (#53) from feat/apply-changes-assign-milestone into main
Reviewed-on: #53
2026-07-09 19:23:49 +00:00
Croissant Le Doux
89c873b368 calibration: count same-day closes honestly (#34)
The cold-start surface showed "N/20 closed issues estimated", implying you're
just (20−N) closes away. But calibrationSamples silently drops closed+estimated
issues that closed in 0 working days (same-day closes) — real closes that
structurally can't calibrate. On this repo that's 10 of 24 closes hidden: the
note read 14/20 as if 6 away, when a third of the history will never count.

- core: `calibrationCoverage(issues, timelines, asOf)` → { candidates, usable,
  excludedSameDay }, counting the silently-excluded same-day closes. Pure, tested.
- surface it: CalibrationData gains `excludedSameDay`; backlogCalibration returns
  the coverage; the Runway note and the Calibration screen now say "… · N same-day
  closes can't calibrate" so the thin sample is explained, not just reported.

Verified on christian/commitea: closed=24, usable=14, excludedSameDay=10.
131 core green (incl. new coverage test); core + desktop typecheck; 14 fixture e2e.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-09 15:18:16 -04:00
11 changed files with 226 additions and 10 deletions

View File

@@ -178,6 +178,9 @@ export function CalibrationScreen({ onBack, data }: { onBack: () => void; data?:
{c.active
? 'You are not bad at estimating; you are optimistic in a very stable way. Stable, I can work with.'
: 'Not enough closed history yet — Im forecasting from cold-start priors and widening the cone to stay honest. The curve takes over at 20.'}
{!c.active && c.excludedSameDay > 0
? ` And ${c.excludedSameDay} closed ${c.excludedSameDay === 1 ? 'issue' : 'issues'} closed the same day they were started — 0 working days cant calibrate, so they dont count toward the 20.`
: ''}
</p>
</Card>
</div>

View File

@@ -17,7 +17,7 @@ export function RunwayScreen({
}: {
onOpenCalibration: () => void
onOpenMilestone: (id?: number) => void
calibration?: { n: number; coldStart: boolean }
calibration?: { n: number; coldStart: boolean; excludedSameDay?: number }
milestones?: RunwayMilestone[]
capacity?: CapacityMember[]
}) {
@@ -30,9 +30,11 @@ export function RunwayScreen({
hours: `${capacityPerWorkday(m).toFixed(2)} pd/day`,
}))
: CAPACITY
const excluded = calibration?.excludedSameDay ?? 0
const calibNote = calibration
? calibration.coldStart
? `cold-start priors · ${calibration.n}/20 closed issues estimated`
? `cold-start priors · ${calibration.n}/20 closed issues estimated` +
(excluded > 0 ? ` · ${excluded} same-day close${excluded === 1 ? '' : 's'} cant calibrate` : '')
: `calibrated on ${calibration.n} closed ${calibration.n === 1 ? 'issue' : 'issues'}`
: 'calibrated on 27 closed issues'
return (

View File

@@ -306,7 +306,15 @@ export function AppShell() {
setMilestoneId(id ?? null)
setView('milestone')
}}
calibration={calibration ? { n: calibration.model.n, coldStart: calibration.model.coldStart } : undefined}
calibration={
calibration
? {
n: calibration.model.n,
coldStart: calibration.model.coldStart,
excludedSameDay: calibration.coverage.excludedSameDay,
}
: undefined
}
milestones={runwayMilestones}
capacity={capacityMembers}
/>

View File

@@ -361,11 +361,14 @@ export interface CalibrationData {
scatter: number[][]
fit: number
effect: { raw: string; banded: string; p50: string }
/** Closed+estimated issues that can't calibrate (same-day / 0-day closes). */
excludedSameDay: number
}
export const CALIBRATION: CalibrationData = {
n: 27,
active: true,
excludedSameDay: 0,
labels: [
{ label: 'est/1d', n: 8, median: '1.1d', bias: 8 },
{ label: 'est/2d', n: 9, median: '2.4d', bias: 18 },

View File

@@ -1,6 +1,8 @@
import {
type CalibrationCoverage,
type CalibrationModel,
type CalibrationSample,
calibrationCoverage,
calibrationSamples,
type CapacityMember,
capacityPerWorkday,
@@ -171,10 +173,11 @@ export function backlogCalibration(
issues: GiteaIssue[],
timelines: Timelines = {},
asOf: Date = new Date(),
): { model: CalibrationModel; data: CalibrationData } {
): { model: CalibrationModel; data: CalibrationData; coverage: CalibrationCoverage } {
const samples = calibrationSamples(issues, timelines, asOf)
const model = fitCalibration(samples)
return { model, data: calibrationData(model, samples, issues) }
const coverage = calibrationCoverage(issues, timelines, asOf)
return { model, coverage, data: calibrationData(model, samples, issues, coverage.excludedSameDay) }
}
const pctFromMu = (mu: number) => Math.round((Math.exp(mu) - 1) * 100)
@@ -188,6 +191,7 @@ export function calibrationData(
model: CalibrationModel,
samples: CalibrationSample[],
openIssues: GiteaIssue[],
excludedSameDay = 0,
): CalibrationData {
const labels = PRIOR_BUCKETS.map((b) => {
const inBucket = samples.filter((s) => s.bucket === b)
@@ -227,6 +231,7 @@ export function calibrationData(
scatter: samples.map((s) => [s.estimateDays, s.actualWorkingDays]),
fit: Number(Math.exp(model.global.mu).toFixed(2)),
effect,
excludedSameDay,
}
}

View File

@@ -17,10 +17,11 @@ export const QUERY_PROJECT_TOOL: ToolDecl = {
properties: {
view: {
type: 'string',
enum: ['focus', 'board', 'calibration', 'issue', 'search'],
enum: ['focus', 'board', 'calibration', 'issue', 'search', 'standup'],
description:
'focus = Now/Next/Later; board = issues by lifecycle column; calibration = estimate-vs-actual; ' +
'issue = one issue (needs filters.issueId); search = issues matching filters.query.',
'issue = one issue (needs filters.issueId); search = issues matching filters.query; ' +
"standup = today's plan + overnight drift + the stalest blocker.",
},
filters: {
type: 'object',

View File

@@ -0,0 +1,74 @@
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 { buildProjectView, type ProjectSnapshot } from './query-project.js'
const asOf = new Date('2026-02-12T00:00:00Z')
function issue(over: Partial<GiteaIssue>): GiteaIssue {
const labels = over.labels ?? ['est/2d']
return {
number: 1,
title: '#1',
body: '',
state: 'open',
labels,
facts: extractLabelFacts(labels),
milestone: null,
assignee: null,
assignees: [],
createdAt: '2026-02-02T09:00:00Z',
updatedAt: '2026-02-02T09:00:00Z',
closedAt: null,
url: '',
...over,
}
}
describe('buildProjectView: standup (#28)', () => {
it('surfaces schedule drift + a stale blocker, with a per-person plan', () => {
// #7 started work 6 working days ago (steeping) and blocks #8; est is 2d → past estimate.
const steeping = issue({ number: 7, labels: ['est/2d', 'p/1'], assignee: 'christian' })
const blocked = issue({ number: 8, labels: ['est/3d', 'p/2'], assignee: 'stephen' })
const timelines: Record<number, LifecycleEvent[]> = {
7: [{ type: 'commit', at: '2026-02-04T09:00:00Z' }],
}
const snap: ProjectSnapshot = { issues: [steeping, blocked], timelines, deps: [{ issue: 8, dependsOn: 7 }] }
const v = buildProjectView('standup', undefined, snap, asOf) as {
plan: { who: string; issue: number; why: string }[]
drift: { issue: number; note: string }[]
nag: { issue: number; steepingDays: number; blocks: number[] } | null
}
// a stale blocker is nagged, and it's the steeping one that blocks another
expect(v.nag).not.toBeNull()
expect(v.nag!.issue).toBe(7)
expect(v.nag!.steepingDays).toBeGreaterThan(0)
expect(v.nag!.blocks).toContain(8)
// drift caught the past-estimate steep
expect(v.drift.some((d) => d.issue === 7)).toBe(true)
// plan gives an earliest pick per person (both assignees represented)
expect(v.plan.map((p) => p.who)).toEqual(expect.arrayContaining(['christian', 'stephen']))
// #7 leads its person's plan on the critical path
expect(v.plan.find((p) => p.issue === 7)?.why).toContain('critical')
})
it('is calm when nothing is steeping (no nag, empty drift)', () => {
const snap: ProjectSnapshot = {
issues: [issue({ number: 1, labels: ['est/2d'] })],
timelines: {},
deps: [],
}
const v = buildProjectView('standup', undefined, snap, asOf) as {
drift: unknown[]
nag: unknown | null
}
expect(v.nag).toBeNull()
expect(v.drift).toEqual([])
})
})

View File

@@ -4,9 +4,9 @@
* deterministic code (scheduler, lifecycle inference, calibration); the model
* only requests a shape and narrates it — it never computes (decisions.md).
*
* v0 serves focus / board / calibration / issue. The remaining views
* (milestone / runway / standup / search) return a `notImplemented` marker so
* the model degrades honestly instead of inventing data.
* Serves focus / board / calibration / issue / search / standup. The remaining
* views (milestone / runway) return a `notImplemented` marker so the model
* degrades honestly instead of inventing data.
*/
import { fitCalibration, calibrationSamples } from '../calibration/calibration-v0.js'
@@ -121,6 +121,62 @@ function searchView(snap: ProjectSnapshot, filters: QueryFilters) {
}
}
/**
* Standup — the morning ritual as a compact, model-narratable payload: today's
* plan (the scheduler's earliest pick per person), overnight drift (real
* anomalies — issues sitting in review or steeping past their estimate), and the
* single stalest blocker to nag about. All from deterministic code; the model
* narrates it. Satisfies #28's "surfaces schedule drift + at least one stale blocker".
*/
function standupView(snap: ProjectSnapshot, asOf: Date) {
const plan = schedule(toSchedulable(snap.issues), snap.deps)
const open = snap.issues.filter((i) => i.state === 'open')
const infOf = (i: GiteaIssue) => inferLifecycle(i, snap.timelines[i.number] ?? [], asOf)
// plan: the earliest scheduled pick per assignee (dependency + priority order)
const seen = new Set<string>()
const planPicks: { who: string; issue: number; title: string; why: string }[] = []
for (const item of plan.items) {
const who = open.find((i) => i.number === item.number)?.assignee ?? 'unassigned'
if (seen.has(who)) continue
seen.add(who)
planPicks.push({
who,
issue: item.number,
title: item.title,
why: item.critical
? 'on the critical path'
: item.blocks.length
? `blocks ${item.blocks.map((b) => `#${b}`).join(', ')}`
: 'next by dependency + priority',
})
}
// drift: real overnight anomalies (review-sitting, steeping past estimate)
const drift: { issue: number; note: string }[] = []
for (const i of open) {
const inf = infOf(i)
if (inf.column === 'review') drift.push({ issue: i.number, note: `#${i.number} is sitting in review` })
else if (inf.steepingDays != null && inf.steepingDays > (i.facts.estimateDays ?? 2))
drift.push({
issue: i.number,
note: `#${i.number} has steeped ${inf.steepingDays}d past its ${i.facts.estimateDays ?? 2}d estimate`,
})
}
// nag: the single longest-steeping open issue + what it blocks
let nag: { issue: number; steepingDays: number; blocks: number[] } | null = null
for (const i of open) {
const inf = infOf(i)
if (inf.steepingDays == null) continue
if (!nag || inf.steepingDays > nag.steepingDays) {
nag = { issue: i.number, steepingDays: inf.steepingDays, blocks: plan.items.find((it) => it.number === i.number)?.blocks ?? [] }
}
}
return { date: asOf.toISOString().slice(0, 10), plan: planPicks.slice(0, 5), drift: drift.slice(0, 5), nag }
}
/** Build the compact payload for one view. Unknown/unbuilt views return a marker. */
export function buildProjectView(
view: ProjectView,
@@ -140,6 +196,8 @@ export function buildProjectView(
return issueView(snap, f, asOf)
case 'search':
return searchView(snap, f)
case 'standup':
return standupView(snap, asOf)
default:
return { notImplemented: view }
}

View File

@@ -5,6 +5,7 @@ import type { LifecycleEvent } from '../lifecycle/lifecycle-v0.js'
import type { GiteaIssue } from '../gitea/types.js'
import {
CALIBRATION_BUCKET_FLOOR,
calibrationCoverage,
calibrationSamples,
type CalibrationSample,
COLD_START_THRESHOLD,
@@ -118,4 +119,26 @@ describe('calibrationSamples', () => {
const noEst = issue({ number: 9, labels: [] })
expect(calibrationSamples([open, noEst], { ...events(8), ...events(9) }, asOf)).toEqual([])
})
it('coverage counts same-day closes as excluded candidates, not as "more closes needed"', () => {
// usable: commit Wed 01-07 → close Mon 01-12 = 3 working days
const usable = issue({ number: 7, labels: ['est/2d'] })
// same-day close: commit and close on the same day = 0 working days → excluded
const sameDay = issue({ number: 10, labels: ['est/2d'], createdAt: '2026-01-12T08:00:00Z' })
const sameDayEvents = {
10: [
{ type: 'commit', at: '2026-01-12T09:00:00Z' } as LifecycleEvent,
{ type: 'close', at: '2026-01-12T17:00:00Z' } as LifecycleEvent,
],
}
const open = issue({ number: 8, state: 'open', labels: ['est/2d'], closedAt: null })
const noEst = issue({ number: 9, labels: [] })
const cov = calibrationCoverage([usable, sameDay, open, noEst], { ...events(7), ...sameDayEvents }, asOf)
expect(cov.candidates).toBe(2) // closed + estimated only (usable + sameDay)
expect(cov.usable).toBe(1)
expect(cov.excludedSameDay).toBe(1)
// the honest denominator: usable matches the model's n
expect(cov.usable).toBe(calibrationSamples([usable, sameDay, open, noEst], { ...events(7), ...sameDayEvents }, asOf).length)
})
})

View File

@@ -125,3 +125,40 @@ export function calibrationSamples(
}
return out
}
/** How the closed+estimated backlog splits into usable samples vs. what can't calibrate. */
export interface CalibrationCoverage {
/** Closed issues carrying an estimate — the calibration candidates. */
candidates: number
/** Candidates that yielded a usable actual (> 0 working days) → become samples. */
usable: number
/**
* Candidates excluded because the issue closed with 0 working days (same-day
* close) or no resolvable actual — real closes that structurally can't
* calibrate. Counting them keeps `usable/threshold` honest: it's not "N more
* closes away" if some of your closes will never count.
*/
excludedSameDay: number
}
/**
* Coverage of the calibration candidates — how many closed+estimated issues are
* usable vs. silently unusable (same-day / 0-day closes). {@link calibrationSamples}
* drops the latter; this counts them so the UI can say *why* the sample is thin.
*/
export function calibrationCoverage(
issues: GiteaIssue[],
timelines: Record<number, LifecycleEvent[]>,
asOf: Date,
): CalibrationCoverage {
let candidates = 0
let usable = 0
for (const issue of issues) {
if (issue.state !== 'closed') continue
if (issue.facts.estimateDays == null) continue
candidates++
const inf = inferLifecycle(issue, timelines[issue.number] ?? [], asOf)
if (inf.actualWorkingDays != null && inf.actualWorkingDays > 0) usable++
}
return { candidates, usable, excludedSameDay: candidates - usable }
}

View File

@@ -80,6 +80,7 @@ export type {
export {
CALIBRATION_BUCKET_FLOOR,
calibrationCoverage,
calibrationSamples,
COLD_START_THRESHOLD,
fitCalibration,
@@ -87,6 +88,7 @@ export {
} from './calibration/calibration-v0.js'
export type {
BucketFit,
CalibrationCoverage,
CalibrationModel,
CalibrationSample,
PersonBias,