Compare commits

11 Commits

Author SHA1 Message Date
Croissant Le Doux
469c63fb95 test: verify record_directive end-to-end + label the ledger in the activity line
The pm-state repo now exists, so the last unverified path is proven: Reginald
records a directive to the real ledger.

- chat panel: the "consulted" activity line names record_directive as "the
  directive ledger" (matching query_project → the project, propose_change → the labels).
- live-directive e2e (gated): a directive typed in chat → Reginald logs it and
  says "recorded", consulting the directive ledger.

Verified live against gemma-4-26b + the commitea-pm-state repo: "freeze scope for
beta, pilots come first" appended to directives/log.jsonl as
{kind: scope, quote: <verbatim>, status: accepted} — logged, never claimed applied.
Reginald is now complete and every tool is verified end-to-end.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 23:07:21 -04:00
3d8a320a04 Merge pull request 'P4: complete Reginald — capture_work + record_directive' (#45) from p4/capture-work into main
Reviewed-on: #45
2026-07-09 03:01:51 +00:00
Croissant Le Doux
ba9ea43b4c feat: record_directive — the PM's ledger in pm-state (P4, completes Reginald)
The last agent tool. When the PM states standing intent ("pilots come first"),
Reginald logs it verbatim to an append-only JSONL ledger in the pm-state repo —
a directive is intent; its effects still land through propose_change. This
completes Reginald's tool surface: query_project · propose_change · capture_work
· record_directive.

core (@commitea/core):
- directives/record-directive-v0: schema (kind/quote/target/params/rationale +
  id/ts/status), serialize/parseDirectiveLog (ts-ordered, seq computed on read,
  corrupt lines skipped), appendDirective (concatenation merge), toDirectiveInput.
- RECORD_DIRECTIVE_TOOL + system prompt update ("log standing intent; never claim
  a change is applied").
- gitea client: getFile/putFile (contents API, base64-agnostic) for the pm-state repo.

app:
- main: a pm-state client (same token, `commitea-pm-state` repo — the purity
  split, D4); appendDirectiveEntry (read→append→write, id/ts stamped here),
  readDirectives. model:chat executes record_directive; pmstate:directives reads
  the ledger. Degrades cleanly when the pm-state repo is absent.
- Directives screen shows the real ledger when present, the fixture demo otherwise.

Note: the pm-state repo isn't created yet — my token lacks write:user (repo
creation). Create `commitea-pm-state` (private) to activate the live path; all the
code + tests are in place. Override with COMMITEA_PMSTATE_REPO.

Verified: 116 core tests green (8 directive + 2 contents-API added), desktop
typecheck clean, 14 fixture e2e green. Gated live test: the real gemma-4-26b calls
record_directive for "pilots come first" (logs intent, doesn't claim to apply it);
the append/read + POST/PUT contents paths are unit-tested.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 22:39:12 -04:00
Croissant Le Doux
6e8a6a15bc feat: capture_work — braindump → decomposed issues → filed in gitea (P4)
The last big agent capability. In the Capture screen, a rough braindump runs real
big-model decomposition into a small, estimated issue set; you review/edit the
labels and approve, and the issues are opened in gitea. This is the one place the
big model earns its keep (docs/agent-tools.md).

core (@commitea/core):
- capture-work: PROPOSE_ISSUES_TOOL + CAPTURE_SYSTEM; captureWork(complete, dump)
  forces a single structured decomposition and returns validated issues; parseCaptureArgs
  drops blank titles + invalid est/p labels. ProposedIssue / CaptureProposal.
- gitea client: createIssue({title, body?, labelIds?}) → POST /issues, normalized.

app:
- model bridge model:capture runs captureWork on the (loaded) big model.
- gitea bridge gitea:createIssues opens each approved issue with its est/* + p/*
  labels (reusing the #41 label-id resolver — zero-pollution, no invented labels).
- Capture screen: when a model is configured, "Brew tickets" runs real capture and
  "Approve all" files the set; otherwise the scripted demo interview runs. Fixed a
  race — the brew handler re-checks model status at click time so a configured
  model never falls into the scripted path before status resolves.

Verified: 108 core tests green (7 capture + createIssue added), desktop typecheck
clean, 14 fixture e2e green. Gated live e2e against gemma-4-26b: the auth braindump
→ 3 real tickets ("Resolve token refresh + session staleness" est/3d p/1, "Fix
webhook double-firing" est/2d p/2, "Write auth setup docs" est/1d p/3), reviewable
and editable; Discard so the test files nothing (createIssue POST is unit-tested).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 22:20:00 -04:00
652c8c86b9 Merge pull request 'P4: writes via chat — Reginald proposes, you approve inline' (#43) from p4/chat-writes into main
Reviewed-on: #43
2026-07-09 02:17:46 +00:00
Croissant Le Doux
47c45ffa3e feat: writes via chat — Reginald proposes, you approve inline (P4)
Completes "chat is the write-path" (decisions.md D1). Ask Reginald to re-estimate
or reprioritize an issue; it formulates a proposal, you approve it inline, and the
write runs through the same guarded apply_changes engine the Issue screen uses.
The model never writes — it proposes; the app owns approval + execution.

core (@commitea/core):
- propose_change tool declaration + REGINALD_SYSTEM updated ("never claim a change
  is applied; you propose, the human approves").
- proposalsFor(args, currentLabels, title): pure — builds the concrete, non-noop
  ChangeProposal(s) (change + label diff) for a propose_change request, dropping
  invalid/unchanged axes. ChangeProposal / ProposeChangeArgs types.

app:
- model bridge executes propose_change by planning against the issue's current
  labels (no write) and returns the proposals with the turn.
- useChat surfaces pending proposals + approve/dismiss; approve calls onApplyChange
  (AppShell's guarded handler → PUT + board/forecast refetch), dismiss leaves it.
- ChatPanel renders each proposal as a propose-approve card (diff + Approve/Dismiss).

Verified: 101 core tests green (4 proposalsFor added), desktop typecheck clean,
14 fixture e2e green. Gated live e2e against gemma-4-26b: "Set the estimate on #3
to est/5d" → Reginald proposes "est/2d → est/5d" as an inline card, says it's
*proposed* not done; Dismiss leaves the repo untouched. The approve→write path is
the #41 engine (separately verified change→revert).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 21:38:25 -04:00
4fc2e83cd5 Merge pull request 'P4: Reginald is real — model router + agent loop + query_project' (#42) from p4/reginald into main
Reviewed-on: #42
2026-07-09 01:28:40 +00:00
Croissant Le Doux
25cf0a6d39 feat: Reginald follows the model you load (auto-detect) + shows it in the header
Instead of a hardcoded model name (which forces LM Studio to JIT-swap your loaded
model out — and fails when a big model already fills memory), resolve the model
at request time: an explicit env override wins, else ask the server which model
is *loaded* (LM Studio's native /api/v0/models), else the first non-embedding
model, else a default. Reginald now uses whatever you load, no config churn.

- main/model.ts: resolveLoadedModel() drives both model:status and model:chat;
  COMMITEA_MODEL_SMALL still overrides.
- useChat exposes the resolved model id; the panel header shows it
  (google/gemma-4-26b-a4b-qat → "gemma-4-26b-a4b · local").
- live-reginald e2e: header assertion relaxed to the loaded model; timeouts
  raised for a slow big local model (~2 calls/turn + a reconcile).

Verified: 14 fixture e2e green; live e2e drives the app against the loaded
gemma-4-26b — "What now?" → "You should work on #2 … on the critical path,
unblocks #33 and #4" (the real scheduler pick), header shows the live model.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 21:23:14 -04:00
Croissant Le Doux
3de887417c feat: Reginald is real — model router + agent loop + query_project (P4)
The fixture chat panel is now a working agent. Ask Reginald a question and it
consults the real project through a tool loop, then answers in grounded prose.
Read-only v0 — writes still go through the propose-approve controls.

core (@commitea/core/agent):
- chat-client: OpenAI-wire chat completions over an injected fetch (same seam as
  gitea). Points at any OpenAI-compatible endpoint (LM Studio/Ollama/OpenAI).
- model-router: small model for prose + the read tool; big model reserved for
  later decomposition (pickModel).
- agent-loop: runAgentTurn drives call→tool→result→call until prose (or a step
  budget), recording each tool step. Injected complete + execute → fully testable.
- query-project: the single read tool's engine — compact focus/board/calibration/
  issue/search views built from scheduler + lifecycle + calibration; unbuilt views
  return a notImplemented marker (never fabricated). The model reports, never computes.
- agent-tools: query_project declaration + Reginald's system prompt.

app:
- main model bridge (model:status, model:chat) runs the loop; query_project
  reconciles the repo and builds the view. Model traffic stays in main (token/CSP).
  gitea.ts refactored to share getGiteaClient + reconcileSnapshot.
- preload + global.d.ts expose the model bridge; useChat drives the panel — real
  agent turn when a model is configured, scripted fixture reply otherwise (so
  fixture e2e is unchanged). A subtle "consulted the project" activity line.

Model config (env, defaults to LM Studio on :1234): COMMITEA_MODEL_URL /
_SMALL (google/gemma-4-e4b) / _BIG (qwen/qwen3.6-35b-a3b). COMMITEA_E2E=1 keeps
it unconfigured so the panel stays scripted.

Verified: 88 core tests green (14 agent: client parse, loop tool/error/budget,
all views) + a gated live integration test. Desktop typecheck clean, 14 fixture
e2e green. Gated live e2e drives the real app against gitea + gemma-4-e4b: asked
"what now?", Reginald called query_project and answered "focus is on issue #2"
(the real scheduler pick).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 21:23:14 -04:00
e8bf71e970 Merge pull request 'P4-3: the write path — apply estimate/priority changes to gitea (apply_changes)' (#41) from p4/apply-changes into main
Reviewed-on: #41
2026-07-09 00:25:26 +00:00
Croissant Le Doux
6198f21d9a feat: the write path — apply estimate/priority changes to gitea (P4-3 apply_changes)
The first write path. Read, forecast, and calibration were all real; now you can
*manage* CommiTea from CommiTea. Estimates/priority are exclusive label axes, so
a change is a label swap — proposed, approved, then written. Nothing is assumed.

core (@commitea/core):
- planIssueChange(current, change): pure diff planner — swaps the est/*|p/* axis,
  clears on null, dedups a doubled axis; returns the resulting label set + a
  before/after diff + noop flag. describeChange() renders "est/2d → est/5d".
- request() seam extended for writes (method/body, JSON, 204). client gains
  listLabels() (name→id) and setIssueLabels() (PUT /issues/{n}/labels).

app:
- main bridge gitea:applyChange — resolves plan.labels → ids (cached, refetch on
  miss), PUTs, returns the plan + fresh issue. Token never leaves main.
- preload + global.d.ts expose applyChange; useBacklog returns a refetch so a
  write re-reconciles the board + forecast.
- Issue screen: an Adjust button (shown only when configured) opens a
  propose-approve Dialog — estimate/priority pickers, live "est/3d → est/8d"
  consequence, Apply/Cancel. AppShell wires it, reflects new labels on the open
  issue immediately, and refetches.

Verified: 83 core tests green (7 apply-changes + 2 client-write new), desktop
typecheck clean, 14 fixture e2e green. Live spec exercises propose + CANCEL (no
mutation); the real PUT was verified once manually (change #2 est/3d→est/8d→200,
reverted clean). Icon: pencil (no sliders-horizontal in the set).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 20:14:47 -04:00
27 changed files with 1452 additions and 63 deletions

View File

@@ -45,6 +45,21 @@ test.describe('live backlog', () => {
await expect(win.getByText(/cold-start · \d+\/20|curve active · n ≥ 20/)).toBeVisible()
await win.screenshot({ path: join(here, '.artifacts', 'screens', 'live-calibration.png'), fullPage: true, animations: 'disabled' })
// Write path (apply_changes): open the top scheduled issue, propose an
// estimate change, and confirm the propose-approve diff renders — then CANCEL
// so the live run never mutates the real repo (the PUT is unit-tested; a
// one-off change→revert verified it against gitea manually).
await rail.getByRole('button', { name: 'Morning service' }).click()
await win.getByRole('main').getByRole('link').first().click()
await expect(win.getByText(/· stephen\/commitea/)).toBeVisible()
await win.getByRole('button', { name: 'Adjust' }).click()
await expect(win.getByText('Adjust estimate & priority')).toBeVisible()
await win.getByRole('combobox').first().selectOption('est/8d')
await expect(win.getByText('Proposed label change')).toBeVisible()
await expect(win.getByText(/est\/8d/).last()).toBeVisible()
await win.screenshot({ path: join(here, '.artifacts', 'screens', 'live-apply-change.png'), fullPage: true, animations: 'disabled' })
await win.getByRole('button', { name: 'Cancel' }).click() // no mutation
await app.close()
})
})

View File

@@ -0,0 +1,37 @@
import { dirname, join } from 'node:path'
import { fileURLToPath } from 'node:url'
import { _electron as electron, expect, test } from '@playwright/test'
const here = dirname(fileURLToPath(import.meta.url))
const MAIN = join(here, '..', 'out', 'main', 'index.js')
// Opt-in (GITEA_LIVE=1 + COMMITEA_MODEL_LIVE=1 + a local model on :1234). Launches
// WITHOUT COMMITEA_E2E so capture_work runs the real decomposition. It Discards at
// the end so the run never files junk issues (the create-issue POST is unit-tested).
test.describe('live capture_work', () => {
test('decomposes a braindump into a reviewable ticket set', async () => {
test.skip(!process.env.GITEA_LIVE || !process.env.COMMITEA_MODEL_LIVE, 'live model test — opt-in')
test.setTimeout(300_000)
const app = await electron.launch({ args: [MAIN], env: { ...process.env } })
const win = await app.firstWindow()
await win.waitForLoadState('domcontentloaded')
const rail = win.getByRole('navigation', { name: 'Primary' })
await rail.getByRole('button', { name: 'Capture' }).click()
await expect(win.getByRole('heading', { name: 'Capture' })).toBeVisible()
// the braindump is prefilled; run the real decomposition
await win.getByRole('button', { name: 'Brew tickets' }).click()
// capture_work → the review tray with a real, estimated ticket set
await expect(win.getByText(/tickets · ~\d+d of work/)).toBeVisible({ timeout: 240_000 })
await expect(win.getByRole('button', { name: 'Approve all' })).toBeVisible()
await win.screenshot({ path: join(here, '.artifacts', 'screens', 'live-capture.png'), fullPage: true, animations: 'disabled' })
// Discard — never file junk issues into the real repo from a test
await win.getByRole('button', { name: 'Discard' }).click()
await expect(win.getByRole('button', { name: 'Brew tickets' })).toBeVisible()
await app.close()
})
})

View File

@@ -0,0 +1,32 @@
import { dirname, join } from 'node:path'
import { fileURLToPath } from 'node:url'
import { _electron as electron, expect, test } from '@playwright/test'
const here = dirname(fileURLToPath(import.meta.url))
const MAIN = join(here, '..', 'out', 'main', 'index.js')
// Opt-in (GITEA_LIVE=1 + COMMITEA_MODEL_LIVE=1 + a local model + the commitea-pm-state
// repo). Reginald records a standing directive; it's appended to the real pm-state
// ledger. The append is idempotent-ish (a fresh line each run); this only reads back
// that a directive was consulted, and leaves the ledger intact.
test.describe('live record_directive', () => {
test('logs a standing directive to the pm-state ledger', async () => {
test.skip(!process.env.GITEA_LIVE || !process.env.COMMITEA_MODEL_LIVE, 'live model test — opt-in')
test.setTimeout(300_000)
const app = await electron.launch({ args: [MAIN], env: { ...process.env } })
const win = await app.firstWindow()
await win.waitForLoadState('domcontentloaded')
await expect(win.getByText(/· local$/)).toBeVisible({ timeout: 20000 })
const composer = win.getByPlaceholder(/Tell me what to do/)
await composer.fill('Record a standing directive: freeze scope for beta, pilots come first.')
await composer.press('Enter')
// the agent logged it to the ledger (record_directive), not applied a change
await expect(win.getByText(/consulted the directive ledger/)).toBeVisible({ timeout: 240_000 })
await win.screenshot({ path: join(here, '.artifacts', 'screens', 'live-directive.png'), fullPage: true, animations: 'disabled' })
await app.close()
})
})

View File

@@ -30,4 +30,27 @@ test.describe('live Reginald', () => {
await app.close()
})
test('proposes an estimate change for inline approval (writes via chat)', async () => {
test.skip(!process.env.GITEA_LIVE || !process.env.COMMITEA_MODEL_LIVE, 'live model test — opt-in')
test.setTimeout(300_000)
const app = await electron.launch({ args: [MAIN], env: { ...process.env } })
const win = await app.firstWindow()
await win.waitForLoadState('domcontentloaded')
await expect(win.getByText(/· local$/)).toBeVisible({ timeout: 20000 })
const composer = win.getByPlaceholder(/Tell me what to do/)
await composer.fill('Set the estimate on issue #3 to est/5d.')
await composer.press('Enter')
// propose_change → an inline propose-approve card (never an auto-write)
await expect(win.getByText('Proposed · #3')).toBeVisible({ timeout: 240_000 })
await expect(win.getByText(/→ est\/5d/)).toBeVisible()
await win.screenshot({ path: join(here, '.artifacts', 'screens', 'live-reginald-propose.png'), fullPage: true, animations: 'disabled' })
// Dismiss so the live run never mutates the repo (the write path itself is #41-tested)
await win.getByRole('button', { name: 'Dismiss' }).click()
await expect(win.getByText(/Left #3 as it was/)).toBeVisible()
await app.close()
})
})

View File

@@ -6,15 +6,24 @@
* properly later.
*/
import { randomUUID } from 'node:crypto'
import { readFileSync } from 'node:fs'
import { dirname, join } from 'node:path'
import {
appendDirective,
createGiteaClient,
type DirectiveEntry,
type GiteaClient,
type GiteaConfig,
type GiteaLabel,
type IssueChange,
type LifecycleEvent,
makeDirectiveEntry,
parseDirectiveLog,
planIssueChange,
type ProjectSnapshot,
type DirectiveInput,
} from '@commitea/core'
import { ipcMain } from 'electron'
@@ -59,6 +68,45 @@ export function getGiteaClient(): GiteaClient | null {
return sharedClient
}
// The pm-state repo holds machine-derived state (the directive ledger). Same
// token/host as the work repo, a different repo (the purity split, decisions D4).
let pmStateClient: GiteaClient | null | undefined
export function getPmStateClient(): GiteaClient | null {
if (pmStateClient === undefined) {
const config = resolveConfig()
pmStateClient = config
? createGiteaClient({ ...config, repo: process.env.COMMITEA_PMSTATE_REPO ?? 'commitea-pm-state' }, fetch)
: null
}
return pmStateClient
}
const DIRECTIVE_LOG_PATH = 'directives/log.jsonl'
async function readDirectiveLog(client: GiteaClient): Promise<{ text: string; sha: string | null }> {
const file = await client.getFile(DIRECTIVE_LOG_PATH)
if (!file) return { text: '', sha: null }
return { text: Buffer.from(file.contentBase64, 'base64').toString('utf8'), sha: file.sha }
}
/** Record a directive: read the ledger, append, write it back (concatenation merge). */
export async function appendDirectiveEntry(client: GiteaClient, input: DirectiveInput): Promise<DirectiveEntry> {
const entry = makeDirectiveEntry(input, randomUUID(), new Date().toISOString())
const { text, sha } = await readDirectiveLog(client)
const next = appendDirective(text, entry)
await client.putFile(DIRECTIVE_LOG_PATH, {
contentBase64: Buffer.from(next, 'utf8').toString('base64'),
message: `directive: ${entry.kind}`,
sha: sha ?? undefined,
})
return entry
}
export async function readDirectives(client: GiteaClient) {
const { text } = await readDirectiveLog(client)
return parseDirectiveLog(text)
}
/** Full reconcile: issues + milestones + native deps + lifecycle timelines. */
export async function reconcileSnapshot(
client: GiteaClient,
@@ -94,4 +142,59 @@ export function registerGiteaIpc(): void {
if (!client) return null
return client.getIssue(index)
})
// Cached label list for name→id resolution; refreshed on demand if a name misses.
let labelCache: GiteaLabel[] | null = null
async function resolveLabelIds(names: string[]): Promise<number[]> {
if (!client) return []
const lookup = () => new Map(labelCache!.map((l) => [l.name, l.id]))
if (!labelCache) labelCache = await client.listLabels()
let byName = lookup()
if (names.some((n) => !byName.has(n))) {
labelCache = await client.listLabels() // a name we don't know — refetch once
byName = lookup()
}
return names.map((n) => byName.get(n)).filter((id): id is number => id != null)
}
// The write path (apply_changes). Additive label swaps, applied only after the
// renderer's propose-approve. Returns the plan + the freshly-read issue.
ipcMain.handle('gitea:applyChange', async (_event, change: IssueChange) => {
if (!client) return { ok: false as const, reason: 'unconfigured' as const }
const current = await client.getIssue(change.issue)
const plan = planIssueChange(current.labels, change)
if (plan.noop) return { ok: true as const, plan, issue: current }
const ids = await resolveLabelIds(plan.labels)
await client.setIssueLabels(change.issue, ids)
const issue = await client.getIssue(change.issue)
return { ok: true as const, plan, issue }
})
// capture_work filing: open each approved issue with its est/* + p/* labels.
// Only touches the CommiTea label namespaces — no invented labels (zero-pollution).
ipcMain.handle(
'gitea:createIssues',
async (_event, issues: { title: string; body?: string; estimate?: string; priority?: string }[]) => {
if (!client) return { ok: false as const, reason: 'unconfigured' as const }
const created: { number: number; title: string }[] = []
for (const it of issues) {
const names = [it.estimate, it.priority].filter((n): n is string => !!n)
const labelIds = await resolveLabelIds(names)
const issue = await client.createIssue({ title: it.title, body: it.body, labelIds })
created.push({ number: issue.number, title: issue.title })
}
return { ok: true as const, created }
},
)
// Read the directive ledger from the pm-state repo (for the Directives screen).
ipcMain.handle('pmstate:directives', async () => {
const pm = getPmStateClient()
if (!pm) return { ok: false as const, reason: 'unconfigured' as const }
try {
return { ok: true as const, directives: await readDirectives(pm) }
} catch (e) {
return { ok: false as const, reason: 'error' as const, message: e instanceof Error ? e.message : String(e) }
}
})
}

View File

@@ -9,18 +9,24 @@
import {
buildProjectView,
captureWork,
type ChangeProposal,
type ChatMessage,
createChatClient,
describeChange,
type ModelRouter,
type ProjectView,
proposalsFor,
type ProposeChangeArgs,
type QueryFilters,
REGINALD_SYSTEM,
REGINALD_TOOLS,
runAgentTurn,
toDirectiveInput,
} from '@commitea/core'
import { ipcMain } from 'electron'
import { getGiteaClient, reconcileSnapshot } from './gitea.js'
import { appendDirectiveEntry, getGiteaClient, getPmStateClient, reconcileSnapshot } from './gitea.js'
/** Small local model for prose + the read tool; big model reserved for later decomposition. */
function resolveModelRouter(): ModelRouter | null {
@@ -80,13 +86,39 @@ export function registerModelIpc(): void {
const model = await resolveLoadedModel(router.small.baseUrl, router.small.model)
const chat = createChatClient({ ...router.small, model }, fetch)
// Proposals the model formulates this turn; the renderer approves them (the
// write happens through gitea:applyChange, never inside the loop).
const proposals: ChangeProposal[] = []
const execute = async (name: string, args: unknown) => {
if (name !== 'query_project') return { error: `unknown tool: ${name}` }
if (!client) return { error: 'gitea is not configured' }
if (name === 'query_project') {
const snap = await reconcileSnapshot(client)
const a = (args ?? {}) as { view: ProjectView; filters?: QueryFilters }
return buildProjectView(a.view, a.filters, snap, new Date())
}
if (name === 'propose_change') {
const a = (args ?? {}) as ProposeChangeArgs
const issue = await client.getIssue(a.issue).catch(() => null)
if (!issue) return { error: `issue #${a.issue} not found` }
const built = proposalsFor(a, issue.labels, issue.title)
proposals.push(...built)
return built.length
? { proposed: built.map((p) => ({ issue: a.issue, diff: describeChange(p.plan) })) }
: { proposed: [], note: 'no change — already at that value' }
}
if (name === 'record_directive') {
const pm = getPmStateClient()
if (!pm) return { error: 'pm-state is not configured' }
try {
const entry = await appendDirectiveEntry(pm, toDirectiveInput(args))
return { recorded: { kind: entry.kind, quote: entry.quote } }
} catch (e) {
return { error: `could not record — is the pm-state repo created? (${e instanceof Error ? e.message : e})` }
}
}
return { error: `unknown tool: ${name}` }
}
try {
const turn = await runAgentTurn({
@@ -95,7 +127,22 @@ export function registerModelIpc(): void {
tools: REGINALD_TOOLS,
execute,
})
return { ok: true as const, content: turn.content, steps: turn.steps }
return { ok: true as const, content: turn.content, steps: turn.steps, proposals }
} catch (e) {
return { ok: false as const, reason: 'error' as const, message: e instanceof Error ? e.message : String(e) }
}
})
// capture_work — braindump → proposed issue set. The big model does the
// decomposition (with one loaded local model, that's the loaded one). Returns
// a proposal; nothing is filed until the Capture tray approves it.
ipcMain.handle('model:capture', async (_event, braindump: string) => {
if (!router) return { ok: false as const, reason: 'unconfigured' as const }
const model = await resolveLoadedModel(router.big.baseUrl, router.big.model)
const chat = createChatClient({ ...router.big, model }, fetch)
try {
const proposal = await captureWork((m, t) => chat.complete(m, t), braindump)
return { ok: true as const, ...proposal }
} catch (e) {
return { ok: false as const, reason: 'error' as const, message: e instanceof Error ? e.message : String(e) }
}

View File

@@ -9,12 +9,22 @@ const api = {
reconcile: () => ipcRenderer.invoke('gitea:reconcile'),
/** One issue by index, normalized (or null if unconfigured). */
getIssue: (index: number) => ipcRenderer.invoke('gitea:getIssue', index),
/** Apply an estimate/priority change (the write path); resolves to the plan + fresh issue. */
applyChange: (change: unknown) => ipcRenderer.invoke('gitea:applyChange', change),
/** File a set of captured issues with their est/* + p/* labels. */
createIssues: (issues: unknown) => ipcRenderer.invoke('gitea:createIssues', issues),
},
pmstate: {
/** Read the directive ledger from the pm-state repo. */
directives: () => ipcRenderer.invoke('pmstate:directives'),
},
model: {
/** Whether a model endpoint is configured (else the UI keeps the scripted Reginald). */
status: () => ipcRenderer.invoke('model:status'),
/** One agent turn: messages in, Reginald's prose + the tools it consulted out. */
chat: (messages: unknown) => ipcRenderer.invoke('model:chat', messages),
/** Decompose a braindump into a proposed issue set (capture_work). */
capture: (braindump: string) => ipcRenderer.invoke('model:capture', braindump),
},
}

View File

@@ -1,8 +1,12 @@
import React from 'react'
import type { ProposedIssue } from '@commitea/core'
import { Badge, Button, Card, Icon, Select, Tag } from '../ui/index.js'
// Capture interview — braindump → interview → approved ticket set (< 2 min)
// Capture interview — braindump → interview → approved ticket set (< 2 min).
// With a model configured, "Brew tickets" runs real capture_work decomposition
// and "Approve all" files the issues in gitea; otherwise the scripted demo runs.
interface Ticket {
title: string
@@ -30,6 +34,64 @@ export function CaptureScreen({ onDone }: { onDone: () => void }) {
const [webEst, setWebEst] = React.useState<string | null>(null)
const [secs, setSecs] = React.useState(0)
// live capture_work path (real model + gitea writes)
const [live, setLive] = React.useState(false)
const [brewing, setBrewing] = React.useState(false)
const [liveTickets, setLiveTickets] = React.useState<ProposedIssue[] | null>(null)
const [filedCount, setFiledCount] = React.useState(0)
const [error, setError] = React.useState<string | null>(null)
React.useEffect(() => {
let alive = true
window.commitea.model.status().then((s) => { if (alive) setLive(s.configured) }).catch(() => {})
return () => { alive = false }
}, [])
const editTicket = (i: number, field: 'estimate' | 'priority', value: string) =>
setLiveTickets((ts) => (ts ? ts.map((t, j) => (j === i ? { ...t, [field]: value } : t)) : ts))
// Decide live-vs-scripted at click time: re-check status if it hasn't resolved
// yet, so a configured model never falls into the scripted interview by a race.
const onBrew = async () => {
const configured = live || (await window.commitea.model.status().then((s) => s.configured).catch(() => false))
if (configured) {
setLive(true)
brew()
} else {
setStage('interview')
}
}
const brew = () => {
setError(null)
setBrewing(true)
window.commitea.model
.capture(dump)
.then((res) => {
setBrewing(false)
if (res.ok && res.issues.length) {
setLiveTickets(res.issues)
setStage('review')
} else {
setError(res.ok ? 'I could not find concrete work in that — try a bit more detail.' : `Capture failed: ${res.reason === 'error' ? res.message : 'no model'}`)
}
})
.catch(() => { setBrewing(false); setError('I could not reach the model.') })
}
const fileLive = () => {
if (!liveTickets) return
setBrewing(true)
window.commitea.gitea
.createIssues(liveTickets)
.then((res) => {
setBrewing(false)
if (res.ok) { setFiledCount(res.created.length); setStage('filed') }
else setError('Filing failed — gitea is not configured.')
})
.catch(() => { setBrewing(false); setError('Filing failed.') })
}
const running = stage === 'interview' || stage === 'review'
React.useEffect(() => {
if (!running) return
@@ -63,19 +125,23 @@ export function CaptureScreen({ onDone }: { onDone: () => void }) {
else setStage('review')
}
// draft tickets build as the interview progresses
const tickets: Ticket[] = []
// draft tickets build as the interview progresses (scripted demo path)
const scripted: Ticket[] = []
if (split === true) {
tickets.push({ title: 'Token refresh: retry with backoff', est: 'est/2d', p: 'p/2' })
tickets.push({ title: 'Session storage: stale reads on wake', est: 'est/1d', p: 'p/3' })
scripted.push({ title: 'Token refresh: retry with backoff', est: 'est/2d', p: 'p/2' })
scripted.push({ title: 'Session storage: stale reads on wake', est: 'est/1d', p: 'p/3' })
} else if (split === false) {
tickets.push({ title: 'Auth: token refresh + session storage', est: 'est/3d', p: 'p/2' })
scripted.push({ title: 'Auth: token refresh + session storage', est: 'est/3d', p: 'p/2' })
}
if (webEst) tickets.push({ title: 'Webhook debounce: double-fire guard', est: webEst, p: 'p/1', dep: 'blocked by auth work' })
if (stage === 'review' || stage === 'filed') {
tickets.push({ title: 'Docs: auth setup guide', est: 'est/1d', p: 'p/4', byReginald: true })
if (webEst) scripted.push({ title: 'Webhook debounce: double-fire guard', est: webEst, p: 'p/1', dep: 'blocked by auth work' })
if (!live && (stage === 'review' || stage === 'filed')) {
scripted.push({ title: 'Docs: auth setup guide', est: 'est/1d', p: 'p/4', byReginald: true })
}
const totalDays = tickets.reduce((n, t) => n + parseInt(t.est.replace('est/', '')), 0)
// real capture_work output overrides the scripted set when present
const tickets: Ticket[] = liveTickets
? liveTickets.map((i) => ({ title: i.title, est: i.estimate ?? 'est/2d', p: i.priority ?? 'p/3' }))
: scripted
const totalDays = tickets.reduce((n, t) => n + parseInt(t.est.replace('est/', ''), 10), 0)
const estOptions = ['est/1d', 'est/2d', 'est/3d', 'est/5d', 'est/8d'].map((v) => ({ value: v, label: v }))
const pOptions = ['p/1', 'p/2', 'p/3', 'p/4'].map((v) => ({ value: v, label: v }))
@@ -96,8 +162,20 @@ export function CaptureScreen({ onDone }: { onDone: () => void }) {
<div style={{ display: 'flex', alignItems: 'center', gap: 8, flexWrap: 'wrap' }}>
{editable ? (
<>
<Select options={estOptions} defaultValue={t.est} style={{ width: 96 }} />
<Select options={pOptions} defaultValue={t.p} style={{ width: 76 }} />
<Select
options={estOptions}
value={liveTickets ? t.est : undefined}
defaultValue={liveTickets ? undefined : t.est}
onChange={(e) => editTicket(i, 'estimate', e.target.value)}
style={{ width: 96 }}
/>
<Select
options={pOptions}
value={liveTickets ? t.p : undefined}
defaultValue={liveTickets ? undefined : t.p}
onChange={(e) => editTicket(i, 'priority', e.target.value)}
style={{ width: 76 }}
/>
</>
) : (
<>
@@ -144,7 +222,12 @@ export function CaptureScreen({ onDone }: { onDone: () => void }) {
<p style={{ font: 'var(--text-agent)', color: 'var(--ink-2)', margin: '10px 0 14px' }}>
Sentences, fragments, grievances all welcome. I'll sort it into tickets and only ask what I can't infer.
</p>
<Button icon="sparkles" onClick={() => setStage('interview')}>Brew tickets</Button>
<Button icon="sparkles" onClick={() => void onBrew()} disabled={brewing}>
{brewing ? 'Brewing…' : 'Brew tickets'}
</Button>
{error ? (
<p style={{ font: 'var(--text-agent)', color: 'var(--danger)', margin: '10px 0 0' }}>{error}</p>
) : null}
</Card>
) : null}
@@ -174,15 +257,26 @@ export function CaptureScreen({ onDone }: { onDone: () => void }) {
<div style={{ display: 'grid', gridTemplateColumns: '1fr 1.1fr', gap: 14, alignItems: 'start' }}>
<Card overline="Consequence" title={`${tickets.length} tickets · ~${totalDays}d of work`} jade
footer={<>
<Button onClick={() => setStage('filed')}>Approve all</Button>
<Button variant="ghost" onClick={() => { setStage('dump'); setQi(0); setLog([]); setSplit(null); setWebEst(null); setSecs(0); }}>Discard</Button>
<Button onClick={() => (live ? fileLive() : setStage('filed'))} disabled={brewing}>
{brewing ? 'Filing…' : 'Approve all'}
</Button>
<Button variant="ghost" onClick={() => { setStage('dump'); setQi(0); setLog([]); setSplit(null); setWebEst(null); setSecs(0); setLiveTickets(null); setError(null); }}>Discard</Button>
</>}>
{live ? (
<p style={{ font: 'var(--text-agent)', color: 'var(--ink-2)', margin: 0 }}>
{tickets.length} issue{tickets.length === 1 ? '' : 's'} from your braindump, estimated and prioritized.
Adjust the labels, then approve I'll open them in gitea with only est/* and p/* labels.
</p>
) : (
<>
<p style={{ font: 'var(--text-body)', margin: '0 0 8px' }}>
Beta's 80% window moves <span style={{ font: 'var(--text-data)' }}>Mar 312 Mar 514</span>. Capacity absorbs the rest.
</p>
<p style={{ font: 'var(--text-agent)', color: 'var(--ink-2)', margin: 0 }}>
I added the docs ticket you mentioned and wired the dependency. Shall I make it so?
</p>
</>
)}
</Card>
<Tray editable />
</div>
@@ -195,7 +289,7 @@ export function CaptureScreen({ onDone }: { onDone: () => void }) {
<Icon name="circle-check" size={22} /> Filed
</span>
<p style={{ font: 'var(--text-body)', margin: 0 }}>
{tickets.length} issues opened in gitea with <span style={{ font: 'var(--text-data)' }}>est/*</span> and <span style={{ font: 'var(--text-data)' }}>p/*</span> labels — nothing else touched.
{live ? filedCount : tickets.length} issues opened in gitea with <span style={{ font: 'var(--text-data)' }}>est/*</span> and <span style={{ font: 'var(--text-data)' }}>p/*</span> labels nothing else touched.
</p>
<p style={{ font: 'var(--text-agent)', color: 'var(--ink-2)', margin: 0 }}>
Elapsed {clock} under budget. No bot comments, no synthetic issues; your repo remains yours.

View File

@@ -8,6 +8,35 @@ export function DirectivesScreen() {
const [pending, setPending] = React.useState<DirectivePending | null>(DIRECTIVES.pending)
const [entries, setEntries] = React.useState<DirectiveEntry[]>(DIRECTIVES.entries)
// Real ledger from the pm-state repo, when it exists — else the fixture demo.
React.useEffect(() => {
let alive = true
window.commitea.pmstate
.directives()
.then((r) => {
if (!alive || !r.ok || r.directives.length === 0) return
setPending(null)
setEntries(
r.directives
.slice()
.reverse()
.map((d) => ({
seq: d.seq,
who: 'You',
when: new Date(d.ts).toLocaleDateString(),
what: d.quote,
why: d.rationale ?? '',
status: d.status === 'accepted' ? 'applied' : d.status === 'withdrawn' ? 'withdrawn' : 'superseded',
consequence: '',
})),
)
})
.catch(() => {})
return () => {
alive = false
}
}, [])
const resolve = (status: string) => {
setEntries((e) => [{
seq: pending!.seq, who: pending!.who, when: pending!.when, what: pending!.what, why: 'pilot demo on the 14th',

View File

@@ -1,17 +1,33 @@
// Issue detail — human intent (gitea) on the left, machine-derived (pm-state) on the right
import React from 'react'
import React, { useState } from 'react'
import {
describeChange,
type EstimateLabel,
ESTIMATE_LABELS,
type IssueChange,
planIssueChange,
type PriorityLabel,
PRIORITY_LABELS,
} from '@commitea/core'
import { ISSUE_DETAIL, type IssueDetail, type IssueRef } from '../../data/fixtures.js'
import { Badge, Button, Card, Icon, Tag } from '../ui/index.js'
import { Badge, Button, Card, Dialog, Icon, Select, Tag } from '../ui/index.js'
const NONE = '—'
export function IssueScreen({
issue,
onBack,
onOpenIssue,
canWrite = false,
onApplyChange,
}: {
issue: IssueRef
onBack: () => void
onOpenIssue: (issue: IssueRef) => void
canWrite?: boolean
onApplyChange?: (change: IssueChange) => Promise<{ ok: boolean }>
}) {
const det: IssueDetail = ISSUE_DETAIL[issue.id] || {
state: 'triage',
@@ -40,6 +56,44 @@ export function IssueScreen({
} as Record<string, { tone: 'warn' | 'neutral' | 'info' | 'ok'; label: string }>
)[det.state] || { tone: 'neutral', label: det.state }
// ---- propose-approve write path (apply_changes) ----
const labels = issue.labels ?? []
const curEstimate = labels.find((l) => l.startsWith('est/')) ?? ''
const curPriority = labels.find((l) => l.startsWith('p/')) ?? ''
const [adjustOpen, setAdjustOpen] = useState(false)
const [estimate, setEstimate] = useState(curEstimate)
const [priority, setPriority] = useState(curPriority)
const [applying, setApplying] = useState(false)
const openAdjust = () => {
setEstimate(curEstimate)
setPriority(curPriority)
setAdjustOpen(true)
}
// the concrete changes this dialog would apply, one per axis that differs
const pendingChanges: IssueChange[] = []
if (estimate !== curEstimate)
pendingChanges.push({ kind: 'reestimate', issue: issue.id, estimate: (estimate || null) as EstimateLabel | null })
if (priority !== curPriority)
pendingChanges.push({ kind: 'reprioritize', issue: issue.id, priority: (priority || null) as PriorityLabel | null })
const diffs = pendingChanges.map((c) => describeChange(planIssueChange(labels, c)))
const apply = async () => {
if (!onApplyChange || pendingChanges.length === 0) return
setApplying(true)
try {
for (const c of pendingChanges) await onApplyChange(c)
setAdjustOpen(false)
} finally {
setApplying(false)
}
}
const estOptions = [{ value: '', label: NONE }, ...ESTIMATE_LABELS.map((l) => ({ value: l, label: l }))]
const prioOptions = [{ value: '', label: NONE }, ...PRIORITY_LABELS.map((l) => ({ value: l, label: l }))]
return (
<div style={{ display: 'flex', flexDirection: 'column', gap: 16 }}>
{/* breadcrumb + header */}
@@ -65,10 +119,69 @@ export function IssueScreen({
</span>
</div>
</div>
<div style={{ display: 'flex', gap: 8, flexShrink: 0 }}>
{canWrite ? (
<Button variant="secondary" icon="pencil" onClick={openAdjust}>
Adjust
</Button>
) : null}
<Button variant="secondary" icon="arrow-up-right">Open in Gitea</Button>
</div>
</header>
</div>
<Dialog
open={adjustOpen}
onClose={() => setAdjustOpen(false)}
title="Adjust estimate & priority"
footer={
<>
<Button variant="ghost" onClick={() => setAdjustOpen(false)}>
Cancel
</Button>
<Button onClick={apply} disabled={pendingChanges.length === 0 || applying}>
{applying ? 'Applying…' : 'Apply to gitea'}
</Button>
</>
}
>
<div style={{ display: 'flex', gap: 14 }}>
<Select
label="Estimate"
options={estOptions}
value={estimate}
onChange={(e) => setEstimate(e.target.value)}
/>
<Select
label="Priority"
options={prioOptions}
value={priority}
onChange={(e) => setPriority(e.target.value)}
/>
</div>
<div style={{ marginTop: 14, minHeight: 40 }}>
{pendingChanges.length === 0 ? (
<p style={{ font: 'var(--text-agent)', color: 'var(--ink-3)', margin: 0 }}>
No change yet pick a different estimate or priority.
</p>
) : (
<>
<p style={{ font: 'var(--text-overline)', letterSpacing: 'var(--letter-spacing-wide)', textTransform: 'uppercase', color: 'var(--ink-3)', margin: '0 0 6px' }}>
Proposed label change
</p>
{diffs.map((d) => (
<div key={d} style={{ font: '500 13px var(--font-mono)', color: 'var(--ink-1)' }}>
{d}
</div>
))}
<p style={{ font: 'var(--text-agent)', color: 'var(--ink-2)', margin: '8px 0 0' }}>
Writes the label to gitea and re-runs the plan. Nothing else changes.
</p>
</>
)}
</div>
</Dialog>
<div style={{ display: 'grid', gridTemplateColumns: '1fr 300px', gap: 16, alignItems: 'start' }}>
{/* left: human intent */}
<div style={{ display: 'flex', flexDirection: 'column', gap: 16 }}>

View File

@@ -1,6 +1,8 @@
import React, { useEffect, useState } from 'react'
import logoIcon from '../../design/assets/logo-icon.png'
import type { IssueChange } from '@commitea/core'
import type { IssueRef } from '../../data/fixtures.js'
import { backlogCalibration, forecastBacklog, issuesToBoardColumns, scheduleFocus } from '../../lib/backlog.js'
import { useBacklog } from '../../lib/use-backlog.js'
@@ -85,7 +87,7 @@ export function AppShell() {
const [offline, setOffline] = useState(false)
const [issue, setIssue] = useState<IssueRef | null>(null)
const [readIds, setReadIds] = useState<number[]>([])
const backlog = useBacklog()
const [backlog, refetchBacklog] = useBacklog()
const boardColumns =
backlog.status === 'ready' ? issuesToBoardColumns(backlog.issues, backlog.timelines) : undefined
const focus =
@@ -107,6 +109,17 @@ export function AppShell() {
setView('issue')
}
// The write path: apply through the bridge, reflect the new labels on the open
// issue immediately, and re-reconcile so the board + forecast catch up.
const applyChange = async (change: IssueChange) => {
const res = await window.commitea.gitea.applyChange(change)
if (res.ok) {
setIssue((cur) => (cur && cur.id === res.issue.number ? { ...cur, labels: res.issue.labels } : cur))
refetchBacklog()
}
return res
}
const NAV: NavEntry[] = [
{ id: 'standup', label: 'Standup', icon: 'sun' },
{ id: 'focus', label: 'Morning service', icon: 'coffee' },
@@ -206,7 +219,13 @@ export function AppShell() {
return <SettingsScreen dark={dark} setDark={setDark} />
case 'issue':
return issue ? (
<IssueScreen issue={issue} onBack={() => setView(prevView)} onOpenIssue={openIssue} />
<IssueScreen
issue={issue}
onBack={() => setView(prevView)}
onOpenIssue={openIssue}
canWrite={backlog.status === 'ready'}
onApplyChange={applyChange}
/>
) : null
case 'states':
return <StatesScreen onCapture={() => setView('capture')} />
@@ -318,7 +337,7 @@ export function AppShell() {
</div>
</main>
<ChatPanel onOpenDirectives={() => setView('directives')} offline={offline} />
<ChatPanel onOpenDirectives={() => setView('directives')} offline={offline} onApplyChange={applyChange} />
</div>
)
}

View File

@@ -1,21 +1,25 @@
import React, { useEffect, useRef, useState } from 'react'
import { describeChange, type IssueChange } from '@commitea/core'
import { useChat } from '../../lib/use-chat.js'
import { Icon, IconButton } from '../ui/index.js'
import { Button, Icon, IconButton } from '../ui/index.js'
/**
* Reginald's panel — chat is the write-path (decisions.md D1). Wired to the
* model bridge via `useChat`: when a model is configured, sending drives a real
* agent turn (query_project + prose); otherwise it echoes the scripted fixture
* reply so the layout stays real. Writes still go through propose-approve.
* model bridge via `useChat`: a configured model drives a real agent turn
* (query_project + prose, propose_change for edits); otherwise it echoes the
* scripted fixture reply. Proposed changes are approved inline here — the write
* runs through `onApplyChange`, the same guarded handler the Issue screen uses.
*/
export interface ChatPanelProps {
onOpenDirectives?: () => void
offline?: boolean
onApplyChange?: (change: IssueChange) => Promise<{ ok: boolean }>
}
export function ChatPanel({ onOpenDirectives, offline }: ChatPanelProps) {
const { msgs, thinking, live, model, steps, send: sendChat } = useChat()
export function ChatPanel({ onOpenDirectives, offline, onApplyChange }: ChatPanelProps) {
const { msgs, thinking, live, model, steps, proposals, send: sendChat, approve, dismiss } = useChat(onApplyChange)
// shorten "google/gemma-4-26b-a4b-qat" → "gemma-4-26b" for the header chip
const modelLabel = model ? (model.split('/').pop() ?? model).replace(/-(qat|instruct|it|gguf)$/i, '') : 'gemma-4'
const [text, setText] = useState('')
@@ -97,9 +101,36 @@ export function ChatPanel({ onOpenDirectives, offline }: ChatPanelProps) {
{thinking ? <div style={{ font: 'var(--text-agent)', color: 'var(--ink-3)' }}>considering</div> : null}
{!thinking && steps.length ? (
<div style={{ font: 'var(--text-caption)', color: 'var(--ink-3)', display: 'flex', alignItems: 'center', gap: 5 }}>
<Icon name="eye" size={11} /> consulted {Array.from(new Set(steps.map((s) => s.replace('query_project', 'the project')))).join(', ')}
<Icon name="eye" size={11} /> consulted {Array.from(new Set(steps.map((s) => s.replace('query_project', 'the project').replace('propose_change', 'the labels').replace('record_directive', 'the directive ledger')))).join(', ')}
</div>
) : null}
{proposals.map((p) => (
<div
key={`${p.change.issue}:${p.change.kind}`}
style={{
border: '1px solid var(--line-2)',
borderRadius: 'var(--radius-2)',
background: 'var(--paper-0)',
padding: '10px 12px',
display: 'flex',
flexDirection: 'column',
gap: 8,
}}
>
<div style={{ font: 'var(--text-caption)', color: 'var(--ink-3)', textTransform: 'uppercase', letterSpacing: 'var(--letter-spacing-wide)' }}>
Proposed · #{p.change.issue}
</div>
<div style={{ font: '500 13px var(--font-mono)', color: 'var(--ink-1)' }}>{describeChange(p.plan)}</div>
<div style={{ display: 'flex', gap: 8 }}>
<Button size="sm" onClick={() => approve(p)} disabled={offline}>
Approve
</Button>
<Button size="sm" variant="ghost" onClick={() => dismiss(p)}>
Dismiss
</Button>
</div>
</div>
))}
</div>
<div style={{ padding: 14, borderTop: '1px solid var(--line-1)', flexShrink: 0 }}>

View File

@@ -1,12 +1,33 @@
import type {
AgentStep,
ChangeProposal,
ChatMessage,
CaptureProposal,
DependencyEdge,
DirectiveRecord,
GiteaIssue,
GiteaMilestone,
IssueChange,
LabelPlan,
LifecycleEvent,
ProposedIssue,
} from '@commitea/core'
/** The result of a write through the bridge. */
export type ApplyChangeResult =
| { ok: false; reason: 'unconfigured' }
| { ok: true; plan: LabelPlan; issue: GiteaIssue }
/** The result of filing captured issues. */
export type CreateIssuesResult =
| { ok: false; reason: 'unconfigured' }
| { ok: true; created: { number: number; title: string }[] }
/** The result of a capture_work decomposition. */
export type CaptureResult =
| { ok: false; reason: 'unconfigured' | 'error'; message?: string }
| ({ ok: true } & CaptureProposal)
/** The gitea bridge exposed by the preload over IPC (main-process backed). */
export interface GiteaBridge {
status(): Promise<{ configured: boolean; repo: string | null }>
@@ -19,17 +40,30 @@ export interface GiteaBridge {
timelines: Record<number, LifecycleEvent[]>
}>
getIssue(index: number): Promise<GiteaIssue | null>
applyChange(change: IssueChange): Promise<ApplyChangeResult>
createIssues(issues: ProposedIssue[]): Promise<CreateIssuesResult>
}
/** One agent turn's result. */
export type ChatResult =
| { ok: false; reason: 'unconfigured' | 'error'; message?: string }
| { ok: true; content: string; steps: AgentStep[] }
| { ok: true; content: string; steps: AgentStep[]; proposals: ChangeProposal[] }
/** The model bridge (Reginald) exposed by the preload over IPC. */
export interface ModelBridge {
status(): Promise<{ configured: boolean; model: string | null }>
chat(messages: ChatMessage[]): Promise<ChatResult>
capture(braindump: string): Promise<CaptureResult>
}
/** The result of reading the directive ledger. */
export type DirectivesResult =
| { ok: false; reason: 'unconfigured' | 'error'; message?: string }
| { ok: true; directives: DirectiveRecord[] }
/** The pm-state bridge (machine-derived state) exposed by the preload over IPC. */
export interface PmStateBridge {
directives(): Promise<DirectivesResult>
}
declare global {
@@ -38,6 +72,7 @@ declare global {
platform: string
gitea: GiteaBridge
model: ModelBridge
pmstate: PmStateBridge
}
}
}

View File

@@ -1,4 +1,4 @@
import { useEffect, useState } from 'react'
import { useCallback, useEffect, useState } from 'react'
import type { DependencyEdge, GiteaIssue, GiteaMilestone, LifecycleEvent } from '@commitea/core'
@@ -15,12 +15,15 @@ export type BacklogState =
}
/**
* Reconcile the managed repo once on mount, through the main-process bridge.
* `unconfigured` means no token — the UI falls back to demo fixtures. Errors
* (network, bad token) surface as `error`.
* Reconcile the managed repo through the main-process bridge. Runs once on
* mount; the returned `refetch` re-reconciles after a write so the board and
* forecast reflect the change. `unconfigured` means no token — the UI falls
* back to demo fixtures. Errors (network, bad token) surface as `error`.
*/
export function useBacklog(): BacklogState {
export function useBacklog(): [BacklogState, () => void] {
const [state, setState] = useState<BacklogState>({ status: 'loading' })
const [nonce, setNonce] = useState(0)
const refetch = useCallback(() => setNonce((n) => n + 1), [])
useEffect(() => {
let alive = true
@@ -46,7 +49,7 @@ export function useBacklog(): BacklogState {
return () => {
alive = false
}
}, [])
}, [nonce])
return state
return [state, refetch]
}

View File

@@ -1,6 +1,6 @@
import { useCallback, useEffect, useRef, useState } from 'react'
import type { ChatMessage as WireMessage } from '@commitea/core'
import type { ChangeProposal, ChatMessage as WireMessage, IssueChange } from '@commitea/core'
import { CANNED_REPLY, CHAT, type ChatMessage } from '../data/fixtures.js'
@@ -18,7 +18,11 @@ export interface ChatState {
model: string | null
/** Tools Reginald consulted on the last turn (for a subtle activity line). */
steps: string[]
/** Changes Reginald has proposed and is awaiting approval on. */
proposals: ChangeProposal[]
send: (text: string) => void
approve: (p: ChangeProposal) => void
dismiss: (p: ChangeProposal) => void
}
/**
@@ -26,18 +30,23 @@ export interface ChatState {
* turn through the main-process bridge (which runs the tool loop). Otherwise it
* echoes the scripted fixture reply, so the layout stays real with no model and
* fixture e2e is unaffected. The fixture greeting is display-only — only real
* turns (`convo`) are sent to the model as history.
* turns (`convo`) are sent to the model as history. `onApplyChange` performs an
* approved write (the same handler the Issue screen uses — it refetches).
*/
export function useChat(): ChatState {
export function useChat(onApplyChange?: (change: IssueChange) => Promise<{ ok: boolean }>): ChatState {
const [seed, setSeed] = useState<ChatMessage[]>(CHAT)
const [convo, setConvo] = useState<ChatMessage[]>([])
const [thinking, setThinking] = useState(false)
const [live, setLive] = useState(false)
const [model, setModel] = useState<string | null>(null)
const [steps, setSteps] = useState<string[]>([])
const [proposals, setProposals] = useState<ChangeProposal[]>([])
const convoRef = useRef(convo)
convoRef.current = convo
const proposalKey = (p: ChangeProposal) => `${p.change.issue}:${p.change.kind}`
const drop = (p: ChangeProposal) => setProposals((ps) => ps.filter((x) => proposalKey(x) !== proposalKey(p)))
useEffect(() => {
let alive = true
window.commitea.model
@@ -63,6 +72,7 @@ export function useChat(): ChatState {
setConvo(nextConvo)
setThinking(true)
setSteps([])
setProposals([])
if (!live) {
window.setTimeout(() => {
@@ -82,6 +92,7 @@ export function useChat(): ChatState {
setThinking(false)
if (res.ok) {
setSteps(res.steps.map((s) => s.tool))
setProposals(res.proposals)
setConvo((c) => [...c, { from: 'agent', text: res.content || '…' }])
} else {
setConvo((c) => [
@@ -101,5 +112,30 @@ export function useChat(): ChatState {
[live],
)
return { msgs: [...seed, ...convo], thinking, live, model, steps, send }
const approve = useCallback(
(p: ChangeProposal) => {
if (!onApplyChange) return
drop(p)
const label = p.plan.added[0] ?? p.plan.removed[0] ?? 'change'
void onApplyChange(p.change).then((res) => {
setConvo((c) => [
...c,
{
from: 'agent',
text: res.ok
? `Done — #${p.change.issue} is now ${label}. The plan's been re-run.`
: `That didn't take — #${p.change.issue} is unchanged.`,
},
])
})
},
[onApplyChange],
)
const dismiss = useCallback((p: ChangeProposal) => {
drop(p)
setConvo((c) => [...c, { from: 'agent', text: `Left #${p.change.issue} as it was.` }])
}, [])
return { msgs: [...seed, ...convo], thinking, live, model, steps, proposals, send, approve, dismiss }
}

View File

@@ -1,4 +1,4 @@
import { describe, expect, it } from 'vitest'
import { beforeAll, describe, expect, it } from 'vitest'
import { extractLabelFacts } from '../labels/label-schema.js'
import type { FetchLike, GiteaIssue } from '../gitea/types.js'
@@ -10,11 +10,22 @@ import { buildProjectView, type ProjectSnapshot } from './query-project.js'
/**
* Opt-in (COMMITEA_MODEL_LIVE=1). Drives the real chat client + agent loop
* against a local OpenAI-compatible server (LM Studio on :1234 by default),
* proving the model calls query_project and narrates the real result.
* proving the model calls the tools and narrates the real result. The model is
* whatever is loaded (via LM Studio's native API), so it never JIT-swaps.
*/
const LIVE = !!process.env.COMMITEA_MODEL_LIVE
const BASE = process.env.COMMITEA_MODEL_URL ?? 'http://localhost:1234/v1'
const MODEL = process.env.COMMITEA_MODEL_SMALL ?? 'google/gemma-4-e4b'
let MODEL = process.env.COMMITEA_MODEL_SMALL ?? ''
beforeAll(async () => {
if (!LIVE || MODEL) return
const root = BASE.replace(/\/v1\/?$/, '')
const loaded = await fetch(`${root}/api/v0/models`)
.then((r) => (r.ok ? (r.json() as Promise<{ data?: { id: string; state?: string; type?: string }[] }>) : null))
.then((d) => d?.data?.find((m) => m.state === 'loaded' && m.type !== 'embeddings')?.id)
.catch(() => undefined)
MODEL = loaded ?? 'google/gemma-4-e4b'
})
function issue(over: Partial<GiteaIssue>): GiteaIssue {
const labels = over.labels ?? []
@@ -59,4 +70,34 @@ describe('agent loop (live model)', () => {
},
60_000,
)
it.skipIf(!LIVE)(
'records a standing instruction via record_directive',
async () => {
const client = createChatClient({ baseUrl: BASE, model: MODEL }, globalThis.fetch as unknown as FetchLike)
const recorded: unknown[] = []
const turn = await runAgentTurn({
complete: (m, t) => client.complete(m, t),
messages: [
{ role: 'system', content: REGINALD_SYSTEM },
{ role: 'user', content: 'Log this standing directive: pilots come first, everything else waits.' },
],
tools: REGINALD_TOOLS,
execute: async (name, args) => {
if (name === 'record_directive') {
recorded.push(args)
return { recorded: { kind: (args as { kind?: string }).kind ?? 'note' } }
}
return name === 'query_project'
? buildProjectView((args as { view: any }).view, (args as any).filters, SNAP, new Date())
: { error: `unknown tool ${name}` }
},
})
// the model logged the directive rather than trying to apply it
expect(turn.steps.some((s) => s.tool === 'record_directive')).toBe(true)
expect(recorded.length).toBeGreaterThan(0)
},
60_000,
)
})

View File

@@ -35,12 +35,57 @@ export const QUERY_PROJECT_TOOL: ToolDecl = {
},
}
export const REGINALD_TOOLS: ToolDecl[] = [QUERY_PROJECT_TOOL]
export const PROPOSE_CHANGE_TOOL: ToolDecl = {
name: 'propose_change',
description:
"Propose an estimate and/or priority change to an issue. This does NOT apply anything — it shows the " +
'human a diff to approve. Use it whenever the user asks to re-estimate or reprioritize. After calling it, ' +
"tell the user you've *proposed* the change for approval — never say it is done.",
parameters: {
type: 'object',
properties: {
issue: { type: 'number', description: 'the issue number to change' },
estimate: { type: 'string', enum: ['est/1d', 'est/2d', 'est/3d', 'est/5d', 'est/8d'] },
priority: { type: 'string', enum: ['p/1', 'p/2', 'p/3', 'p/4'] },
},
required: ['issue'],
},
}
export const RECORD_DIRECTIVE_TOOL: ToolDecl = {
name: 'record_directive',
description:
'Log a standing instruction from the PM to the durable directive ledger — a reprioritization, ' +
'a re-estimate policy, a deadline, a scope or capacity call, or a plain note. Use it when the user ' +
'states intent that should persist ("pilots come first", "freeze scope for beta"). This records the ' +
'intent verbatim; the actual issue edits still go through propose_change.',
parameters: {
type: 'object',
properties: {
kind: { type: 'string', enum: ['reprioritize', 'reestimate', 'set-deadline', 'scope', 'capacity', 'note'] },
quote: { type: 'string', description: "the PM's own words, stored verbatim" },
target: {
type: 'object',
properties: {
issue: { type: 'number' },
milestone: { type: 'number' },
member: { type: 'string' },
},
},
rationale: { type: 'string', description: 'why (optional)' },
},
required: ['kind', 'quote'],
},
}
export const REGINALD_TOOLS: ToolDecl[] = [QUERY_PROJECT_TOOL, PROPOSE_CHANGE_TOOL, RECORD_DIRECTIVE_TOOL]
export const REGINALD_SYSTEM = [
'You are Reginald, the calm, dry project manager inside CommiTea — a tool that runs projects on Gitea.',
'Call query_project to ground every answer in the real project; never invent issues, numbers, or dates.',
'The scheduler and forecasts are deterministic code — report their output, do not recompute it.',
'Forecasts are ranges, never single dates. Refer to issues as #<number>.',
'Be brief and plain. A sentence or two is usually enough. No preamble, no bullet-point dumps.',
'To change an estimate or priority, call propose_change — it shows the human a diff to approve.',
'When the PM states standing intent ("pilots first", "freeze scope"), call record_directive to log it.',
'Never claim a change is applied; you propose, the human approves. Forecasts are ranges, never single dates.',
'Refer to issues as #<number>. Be brief and plain — a sentence or two. No preamble, no bullet dumps.',
].join(' ')

View File

@@ -0,0 +1,65 @@
import { describe, expect, it } from 'vitest'
import type { CompletionResult } from './chat-client.js'
import { captureWork, parseCaptureArgs } from './capture-work.js'
describe('parseCaptureArgs', () => {
it('validates issues and keeps only real titles + valid labels', () => {
const p = parseCaptureArgs({
issues: [
{ title: 'Fix token refresh', body: 'dies silently', estimate: 'est/2d', priority: 'p/1' },
{ title: ' ', body: 'blank title dropped' },
{ title: 'Docs', estimate: 'est/9d', priority: 'urgent' }, // invalid labels → dropped to undefined
],
consequence: 'Beta slips a day',
})
expect(p.issues).toHaveLength(2)
expect(p.issues[0]).toEqual({ title: 'Fix token refresh', body: 'dies silently', estimate: 'est/2d', priority: 'p/1' })
expect(p.issues[1]).toEqual({ title: 'Docs', body: '', estimate: undefined, priority: undefined })
expect(p.consequence).toBe('Beta slips a day')
})
it('tolerates a missing/!array issues field', () => {
expect(parseCaptureArgs({}).issues).toEqual([])
expect(parseCaptureArgs({ issues: 'nope' }).issues).toEqual([])
})
})
describe('captureWork', () => {
it('forces propose_issues and returns the validated set', async () => {
const result: CompletionResult = {
content: '',
toolCalls: [
{
id: 'c1',
name: 'propose_issues',
arguments: JSON.stringify({
issues: [{ title: 'Retry token refresh with backoff', body: '', estimate: 'est/2d', priority: 'p/2' }],
consequence: 'no material shift',
}),
},
],
}
let sawTool = ''
const proposal = await captureWork(async (_m, tools) => {
sawTool = tools?.[0]?.name ?? ''
return result
}, 'auth is flaky, token refresh dies')
expect(sawTool).toBe('propose_issues')
expect(proposal.issues[0].title).toBe('Retry token refresh with backoff')
expect(proposal.consequence).toBe('no material shift')
})
it('returns an empty set when the model answers without the tool', async () => {
const proposal = await captureWork(async () => ({ content: 'I need more detail.', toolCalls: [] }), 'vague')
expect(proposal.issues).toEqual([])
})
it('survives malformed tool arguments', async () => {
const proposal = await captureWork(
async () => ({ content: '', toolCalls: [{ id: 'c1', name: 'propose_issues', arguments: '{not json' }] }),
'x',
)
expect(proposal.issues).toEqual([])
})
})

View File

@@ -0,0 +1,109 @@
/**
* capture_work — braindump → a small set of concrete issues. This is the one
* place the big model earns its keep (decomposition + estimate negotiation).
* It returns a *proposal*; nothing is filed until the human approves it in the
* Capture tray and it goes through the create-issue write path. Pure
* orchestration over an injected `complete` — stubbable, so it's testable offline.
*/
import {
type EstimateLabel,
ESTIMATE_LABELS,
type PriorityLabel,
PRIORITY_LABELS,
} from '../labels/label-schema.js'
import type { ChatMessage, CompletionResult, ToolDecl } from './chat-client.js'
export interface ProposedIssue {
title: string
body: string
estimate?: EstimateLabel
priority?: PriorityLabel
}
export interface CaptureProposal {
issues: ProposedIssue[]
/** One-line schedule impact, if the model offered one. */
consequence?: string
}
export const PROPOSE_ISSUES_TOOL: ToolDecl = {
name: 'propose_issues',
description: 'Return the decomposed issue set for a braindump. Call this exactly once.',
parameters: {
type: 'object',
properties: {
issues: {
type: 'array',
items: {
type: 'object',
properties: {
title: { type: 'string', description: 'a clear imperative title' },
body: { type: 'string', description: 'one or two lines of detail' },
estimate: { type: 'string', enum: ['est/1d', 'est/2d', 'est/3d', 'est/5d', 'est/8d'] },
priority: { type: 'string', enum: ['p/1', 'p/2', 'p/3', 'p/4'] },
},
required: ['title'],
},
},
consequence: { type: 'string', description: 'one-line note on the schedule impact' },
},
required: ['issues'],
},
}
export const CAPTURE_SYSTEM = [
'You are Reginald, decomposing a rough braindump into a small set of concrete Gitea issues.',
'Call propose_issues exactly once. Split genuinely separate work; merge trivially-coupled work; invent no scope.',
'Each issue gets an imperative title, a one-line body, an estimate (est/1d…8d) and a priority (p/1…4).',
'Estimate honestly — a "quick" task is rarely one day. Keep the set tight; three good issues beat eight vague ones.',
].join(' ')
function isEstimate(v: unknown): v is EstimateLabel {
return typeof v === 'string' && (ESTIMATE_LABELS as readonly string[]).includes(v)
}
function isPriority(v: unknown): v is PriorityLabel {
return typeof v === 'string' && (PRIORITY_LABELS as readonly string[]).includes(v)
}
/** Coerce the model's raw propose_issues args into a validated proposal. */
export function parseCaptureArgs(args: unknown): CaptureProposal {
const a = (args ?? {}) as { issues?: unknown[]; consequence?: unknown }
const issues: ProposedIssue[] = []
for (const raw of Array.isArray(a.issues) ? a.issues : []) {
const r = (raw ?? {}) as Record<string, unknown>
const title = typeof r.title === 'string' ? r.title.trim() : ''
if (!title) continue
issues.push({
title,
body: typeof r.body === 'string' ? r.body : '',
estimate: isEstimate(r.estimate) ? r.estimate : undefined,
priority: isPriority(r.priority) ? r.priority : undefined,
})
}
return { issues, consequence: typeof a.consequence === 'string' ? a.consequence : undefined }
}
/**
* Run one decomposition turn. Forces the model to answer via propose_issues and
* returns the validated set. An empty set means the model declined to structure it.
*/
export async function captureWork(
complete: (messages: ChatMessage[], tools?: ToolDecl[]) => Promise<CompletionResult>,
braindump: string,
): Promise<CaptureProposal> {
const res = await complete(
[
{ role: 'system', content: CAPTURE_SYSTEM },
{ role: 'user', content: braindump },
],
[PROPOSE_ISSUES_TOOL],
)
const call = res.toolCalls.find((t) => t.name === 'propose_issues')
if (!call) return { issues: [] }
try {
return parseCaptureArgs(JSON.parse(call.arguments || '{}'))
} catch {
return { issues: [] }
}
}

View File

@@ -0,0 +1,79 @@
import { describe, expect, it } from 'vitest'
import { describeChange, type IssueChange, planIssueChange, proposalsFor } from './apply-changes-v0.js'
describe('planIssueChange', () => {
it('swaps the estimate label, keeping non-axis labels', () => {
const plan = planIssueChange(['est/2d', 'p/1', 'backend'], { kind: 'reestimate', issue: 1, estimate: 'est/5d' })
expect(plan.removed).toEqual(['est/2d'])
expect(plan.added).toEqual(['est/5d'])
expect(plan.labels).toEqual(['p/1', 'backend', 'est/5d'])
expect(plan.noop).toBe(false)
})
it('adds an estimate when none was set', () => {
const plan = planIssueChange(['p/2'], { kind: 'reestimate', issue: 1, estimate: 'est/1d' })
expect(plan.removed).toEqual([])
expect(plan.added).toEqual(['est/1d'])
expect(plan.labels).toEqual(['p/2', 'est/1d'])
})
it('clears the axis when the target is null', () => {
const plan = planIssueChange(['est/3d', 'p/1'], { kind: 'reprioritize', issue: 1, priority: null })
expect(plan.removed).toEqual(['p/1'])
expect(plan.added).toEqual([])
expect(plan.labels).toEqual(['est/3d'])
})
it('is a noop when the target already holds the axis alone', () => {
const plan = planIssueChange(['est/2d', 'p/1'], { kind: 'reestimate', issue: 1, estimate: 'est/2d' })
expect(plan.noop).toBe(true)
expect(plan.labels).toEqual(['p/1', 'est/2d'])
})
it('cleans up a duplicated axis down to the target', () => {
// two est/* labels — the change collapses to one
const plan = planIssueChange(['est/2d', 'est/5d', 'p/1'], { kind: 'reestimate', issue: 1, estimate: 'est/5d' })
expect(plan.removed).toEqual(['est/2d'])
expect(plan.added).toEqual([]) // est/5d already present
expect(plan.labels).toEqual(['p/1', 'est/5d'])
expect(plan.noop).toBe(false)
})
it('reprioritize only touches the priority axis', () => {
const plan = planIssueChange(['est/2d', 'p/3'], { kind: 'reprioritize', issue: 1, priority: 'p/1' })
expect(plan.labels).toEqual(['est/2d', 'p/1'])
})
it('describeChange renders the diff', () => {
const change: IssueChange = { kind: 'reestimate', issue: 1, estimate: 'est/5d' }
expect(describeChange(planIssueChange(['est/2d'], change))).toBe('est/2d → est/5d')
expect(describeChange(planIssueChange(['p/1'], { kind: 'reprioritize', issue: 1, priority: null }))).toBe('p/1 → ∅')
expect(describeChange(planIssueChange(['est/5d'], change))).toBe('no change')
})
})
describe('proposalsFor', () => {
it('builds one proposal per changed axis, carrying the concrete change + diff', () => {
const props = proposalsFor({ issue: 2, estimate: 'est/5d', priority: 'p/1' }, ['est/2d', 'p/3'], 'ChangeSource')
expect(props).toHaveLength(2)
expect(props[0].change).toEqual({ kind: 'reestimate', issue: 2, estimate: 'est/5d' })
expect(describeChange(props[0].plan)).toBe('est/2d → est/5d')
expect(props[1].change).toEqual({ kind: 'reprioritize', issue: 2, priority: 'p/1' })
expect(props[0].issueTitle).toBe('ChangeSource')
})
it('drops a noop axis (already at the requested value)', () => {
const props = proposalsFor({ issue: 2, estimate: 'est/2d', priority: 'p/1' }, ['est/2d', 'p/3'])
expect(props.map((p) => p.change.kind)).toEqual(['reprioritize']) // estimate unchanged
})
it('ignores invalid label values from the model', () => {
const props = proposalsFor({ issue: 2, estimate: 'est/4d' as never, priority: 'high' as never }, [])
expect(props).toEqual([])
})
it('returns nothing when no axis is provided', () => {
expect(proposalsFor({ issue: 2 }, ['est/2d'])).toEqual([])
})
})

View File

@@ -0,0 +1,104 @@
/**
* Apply-changes, v0 — the write path's planning half. Estimates and priority
* live as exclusive label axes (`est/*`, `p/*`); a change swaps the axis label.
* This computes the resulting label set + a human-readable diff *purely*, so the
* UI can show a propose-approve consequence before the write and tests can pin
* the semantics. The actual PUT (name→id resolution + network) is the bridge's
* job — additive here, never assumed.
*/
import {
type EstimateLabel,
ESTIMATE_LABELS,
type PriorityLabel,
PRIORITY_LABELS,
} from '../labels/label-schema.js'
export type IssueChange =
| { kind: 'reestimate'; issue: number; estimate: EstimateLabel | null }
| { kind: 'reprioritize'; issue: number; priority: PriorityLabel | null }
export interface LabelPlan {
/** The full resulting label-name set (order: kept labels, then the new axis label). */
labels: string[]
/** Axis labels being added (0 or 1). */
added: string[]
/** Axis labels being removed (includes clearing a duplicated axis). */
removed: string[]
/** true when the change would leave the labels unchanged. */
noop: boolean
}
function axisFor(change: IssueChange): { labels: readonly string[]; target: string | null } {
return change.kind === 'reestimate'
? { labels: ESTIMATE_LABELS, target: change.estimate }
: { labels: PRIORITY_LABELS, target: change.priority }
}
/**
* Plan the label mutation for a single change. Removes every label on the
* change's axis except the target, adds the target if absent. Setting the axis
* to null clears it. Cleans up a duplicated axis (two `est/*`) as a side effect.
*/
export function planIssueChange(current: string[], change: IssueChange): LabelPlan {
const { labels: axis, target } = axisFor(change)
const onAxis = current.filter((l) => axis.includes(l))
const removed = onAxis.filter((l) => l !== target)
const added = target && !current.includes(target) ? [target] : []
const kept = current.filter((l) => !axis.includes(l))
const labels = target ? [...kept, target] : kept
return { labels, added, removed, noop: added.length === 0 && removed.length === 0 }
}
/** A short "est/2d → est/3d" (or "+p/1" / "est/5d") summary for the confirm UI. */
export function describeChange(plan: LabelPlan): string {
if (plan.noop) return 'no change'
const from = plan.removed.length ? plan.removed.join(', ') : '∅'
const to = plan.added.length ? plan.added.join(', ') : '∅'
return `${from}${to}`
}
/** A change the agent proposes: the concrete op + its diff, ready for approve-then-apply. */
export interface ChangeProposal {
change: IssueChange
plan: LabelPlan
issueTitle?: string
}
/** What the `propose_change` tool accepts — a target issue and the axes to set. */
export interface ProposeChangeArgs {
issue: number
estimate?: EstimateLabel
priority?: PriorityLabel
}
function isEstimate(v: unknown): v is EstimateLabel {
return typeof v === 'string' && (ESTIMATE_LABELS as readonly string[]).includes(v)
}
function isPriority(v: unknown): v is PriorityLabel {
return typeof v === 'string' && (PRIORITY_LABELS as readonly string[]).includes(v)
}
/**
* Build the concrete, non-noop proposals for a `propose_change` request against
* an issue's current labels. Invalid or unchanged axes are dropped — the agent
* proposes only real changes, and never a label outside the est/* · p/* axes.
*/
export function proposalsFor(
args: ProposeChangeArgs,
currentLabels: string[],
issueTitle?: string,
): ChangeProposal[] {
const out: ChangeProposal[] = []
if (isEstimate(args.estimate)) {
const change: IssueChange = { kind: 'reestimate', issue: args.issue, estimate: args.estimate }
const plan = planIssueChange(currentLabels, change)
if (!plan.noop) out.push({ change, plan, issueTitle })
}
if (isPriority(args.priority)) {
const change: IssueChange = { kind: 'reprioritize', issue: args.issue, priority: args.priority }
const plan = planIssueChange(currentLabels, change)
if (!plan.noop) out.push({ change, plan, issueTitle })
}
return out
}

View File

@@ -0,0 +1,66 @@
import { describe, expect, it } from 'vitest'
import {
appendDirective,
makeDirectiveEntry,
parseDirectiveLog,
serializeDirective,
toDirectiveInput,
} from './record-directive-v0.js'
describe('toDirectiveInput', () => {
it('keeps a valid kind + target and drops an empty target', () => {
const input = toDirectiveInput({ kind: 'reprioritize', quote: 'pilots first', target: { issue: 87 }, rationale: 'blocked' })
expect(input).toEqual({ kind: 'reprioritize', quote: 'pilots first', target: { issue: 87 }, params: undefined, rationale: 'blocked' })
expect(toDirectiveInput({ kind: 'note', quote: 'x', target: {} }).target).toBeUndefined()
})
it('falls back to note for an unknown kind', () => {
expect(toDirectiveInput({ kind: 'nonsense', quote: 'hmm' }).kind).toBe('note')
})
})
describe('serialize + parse round-trip', () => {
const entry = makeDirectiveEntry(
{ kind: 'reprioritize', quote: 'pilots come first', target: { issue: 87 } },
'id-1',
'2026-02-01T09:00:00Z',
)
it('serializes to one JSON line', () => {
const line = serializeDirective(entry)
expect(line).not.toContain('\n')
expect(JSON.parse(line)).toMatchObject({ id: 'id-1', kind: 'reprioritize', status: 'accepted' })
})
it('parses a log, orders by ts, and assigns a 1-based seq', () => {
const a = serializeDirective(makeDirectiveEntry({ kind: 'note', quote: 'later' }, 'b', '2026-02-02T00:00:00Z'))
const b = serializeDirective(makeDirectiveEntry({ kind: 'note', quote: 'earlier' }, 'a', '2026-02-01T00:00:00Z'))
const records = parseDirectiveLog(`${a}\n${b}\n`)
expect(records.map((r) => r.quote)).toEqual(['earlier', 'later'])
expect(records.map((r) => r.seq)).toEqual([1, 2])
})
it('skips blank and corrupt lines without losing the rest', () => {
const good = serializeDirective(entry)
const records = parseDirectiveLog(`\n{not json\n${good}\n\n`)
expect(records).toHaveLength(1)
expect(records[0].id).toBe('id-1')
})
})
describe('appendDirective', () => {
it('concatenates a newline-terminated entry, normalizing a missing trailing newline', () => {
const e1 = makeDirectiveEntry({ kind: 'note', quote: 'one' }, 'i1', '2026-01-01T00:00:00Z')
const e2 = makeDirectiveEntry({ kind: 'note', quote: 'two' }, 'i2', '2026-01-02T00:00:00Z')
let log = appendDirective('', e1)
log = appendDirective(log, e2)
expect(parseDirectiveLog(log).map((r) => r.quote)).toEqual(['one', 'two'])
expect(log.endsWith('\n')).toBe(true)
})
it('handles existing text without a trailing newline', () => {
const e = makeDirectiveEntry({ kind: 'note', quote: 'x' }, 'i', '2026-01-01T00:00:00Z')
expect(appendDirective('{"id":"prev","ts":"2025-01-01T00:00:00Z"}', e).split('\n').filter(Boolean)).toHaveLength(2)
})
})

View File

@@ -0,0 +1,106 @@
/**
* record_directive — the PM's standing instructions ("pilots come first"),
* appended to an append-only JSONL ledger in the pm-state repo (decisions.md D4,
* pm-state.md). A directive is *intent*: it's logged verbatim; its effects land
* later through apply_changes. Merge is concatenation — order derives from `ts`
* at read time, so two writers never conflict. `seq` is a display ordinal
* computed on read, never stored. This module is pure serialize/parse; the
* append (read → concat → write) is the bridge's job.
*/
export type DirectiveKind = 'reprioritize' | 'reestimate' | 'set-deadline' | 'scope' | 'capacity' | 'note'
export type DirectiveStatus = 'proposed' | 'accepted' | 'amended' | 'withdrawn'
export interface DirectiveTarget {
issue?: number
milestone?: number
member?: string
}
/** What the record_directive tool captures. */
export interface DirectiveInput {
kind: DirectiveKind
/** Verbatim PM words, shown in the ledger. */
quote: string
target?: DirectiveTarget
/** Structured effect the scheduler applies, e.g. { priority: 1 }. */
params?: Record<string, unknown>
rationale?: string
}
/** A ledger entry — an input plus its durable id/ts/status. */
export interface DirectiveEntry extends DirectiveInput {
id: string
ts: string
status: DirectiveStatus
}
/** A ledger entry as read back, with a computed display ordinal. */
export interface DirectiveRecord extends DirectiveEntry {
seq: number
}
const DIRECTIVE_KINDS: readonly DirectiveKind[] = [
'reprioritize',
'reestimate',
'set-deadline',
'scope',
'capacity',
'note',
]
/** Normalize a raw tool payload into a DirectiveInput (unknown kind → note). */
export function toDirectiveInput(raw: unknown): DirectiveInput {
const r = (raw ?? {}) as Record<string, unknown>
const kind = DIRECTIVE_KINDS.includes(r.kind as DirectiveKind) ? (r.kind as DirectiveKind) : 'note'
const target = (r.target ?? undefined) as DirectiveTarget | undefined
return {
kind,
quote: typeof r.quote === 'string' ? r.quote : '',
target: target && (target.issue || target.milestone || target.member) ? target : undefined,
params: (r.params && typeof r.params === 'object' ? (r.params as Record<string, unknown>) : undefined),
rationale: typeof r.rationale === 'string' ? r.rationale : undefined,
}
}
/** Build a full entry from an input + externally-supplied id/ts (Date/uuid live in the caller). */
export function makeDirectiveEntry(
input: DirectiveInput,
id: string,
ts: string,
status: DirectiveStatus = 'accepted',
): DirectiveEntry {
return { ...input, id, ts, status }
}
/** One JSONL line (no trailing newline — the caller joins). */
export function serializeDirective(entry: DirectiveEntry): string {
return JSON.stringify(entry)
}
/**
* Parse a JSONL log into records ordered by `ts` (then id for stability), with a
* 1-based `seq` assigned on read. Blank/corrupt lines are skipped, not fatal.
*/
export function parseDirectiveLog(text: string): DirectiveRecord[] {
const entries: DirectiveEntry[] = []
for (const line of text.split('\n')) {
const trimmed = line.trim()
if (!trimmed) continue
try {
const e = JSON.parse(trimmed) as DirectiveEntry
if (e && typeof e.id === 'string' && typeof e.ts === 'string') entries.push(e)
} catch {
// skip a corrupt line rather than lose the whole ledger
}
}
entries.sort((a, b) => (a.ts === b.ts ? a.id.localeCompare(b.id) : a.ts.localeCompare(b.ts)))
return entries.map((e, i) => ({ ...e, seq: i + 1 }))
}
/** Append a serialized entry to existing log text (concatenation merge). */
export function appendDirective(existing: string, entry: DirectiveEntry): string {
const base = existing.endsWith('\n') || existing === '' ? existing : existing + '\n'
return `${base}${serializeDirective(entry)}\n`
}

View File

@@ -123,6 +123,63 @@ describe('createGiteaClient.getIssue', () => {
)
})
it('setIssueLabels PUTs the label ids to the issue labels endpoint', async () => {
const { fetch, calls } = stubFetch(null, 204)
await createGiteaClient(CONFIG, fetch).setIssueLabels(9, [3, 7])
expect(calls).toHaveLength(1)
expect(calls[0].url).toBe('https://gitea.stephenmann.io/api/v1/repos/christian/commitea/issues/9/labels')
expect(calls[0].init?.method).toBe('PUT')
expect(calls[0].init?.headers?.['Content-Type']).toBe('application/json')
expect(JSON.parse(calls[0].init?.body ?? '{}')).toEqual({ labels: [3, 7] })
})
it('listLabels returns id+name pairs', async () => {
const { fetch } = stubFetch([
{ id: 3, name: 'est/2d', color: 'fff' },
{ id: 7, name: 'p/1', color: '000' },
])
const labels = await createGiteaClient(CONFIG, fetch).listLabels()
expect(labels).toEqual([
{ id: 3, name: 'est/2d' },
{ id: 7, name: 'p/1' },
])
})
it('getFile returns null on 404 and content+sha on hit', async () => {
const miss = stubFetch('nope', 404)
expect(await createGiteaClient(CONFIG, miss.fetch).getFile('directives/log.jsonl')).toBeNull()
const hit = stubFetch({ content: 'aGVsbG8=\n', sha: 'abc123' })
const file = await createGiteaClient(CONFIG, hit.fetch).getFile('directives/log.jsonl')
expect(file).toEqual({ contentBase64: 'aGVsbG8=', sha: 'abc123' })
expect(hit.calls[0].url).toContain('/contents/directives/log.jsonl')
})
it('putFile POSTs to create and PUTs to update (with sha)', async () => {
const create = stubFetch({}, 201)
await createGiteaClient(CONFIG, create.fetch).putFile('directives/log.jsonl', { contentBase64: 'eA==', message: 'seed' })
expect(create.calls[0].init?.method).toBe('POST')
expect(JSON.parse(create.calls[0].init?.body ?? '{}')).toEqual({ content: 'eA==', message: 'seed' })
const update = stubFetch({}, 200)
await createGiteaClient(CONFIG, update.fetch).putFile('directives/log.jsonl', { contentBase64: 'eQ==', message: 'append', sha: 's1' })
expect(update.calls[0].init?.method).toBe('PUT')
expect(JSON.parse(update.calls[0].init?.body ?? '{}')).toEqual({ content: 'eQ==', message: 'append', sha: 's1' })
})
it('createIssue POSTs title/body/labels and returns a normalized issue', async () => {
const created = { ...RAW_ISSUE, number: 44, title: 'Retry token refresh', labels: [{ name: 'est/2d' }, { name: 'p/2' }] }
const { fetch, calls } = stubFetch(created, 201)
const issue = await createGiteaClient(CONFIG, fetch).createIssue({ title: 'Retry token refresh', body: 'backoff', labelIds: [3, 7] })
expect(calls[0].url).toBe('https://gitea.stephenmann.io/api/v1/repos/christian/commitea/issues')
expect(calls[0].init?.method).toBe('POST')
expect(JSON.parse(calls[0].init?.body ?? '{}')).toEqual({ title: 'Retry token refresh', body: 'backoff', labels: [3, 7] })
expect(issue.number).toBe(44)
expect(issue.labels).toEqual(['est/2d', 'p/2'])
})
it('throws GiteaApiError carrying status + body on a non-2xx response', async () => {
const { fetch } = stubFetch('not found', 404)
const client = createGiteaClient(CONFIG, fetch)

View File

@@ -13,6 +13,7 @@ import {
type FetchLike,
type GiteaConfig,
type GiteaIssue,
type GiteaLabel,
type GiteaMilestone,
type GiteaMilestoneRef,
} from './types.js'
@@ -98,6 +99,16 @@ export interface GiteaClient {
getIssueDependencies(index: number): Promise<number[]>
/** Normalized lifecycle events for one issue (all pages of its timeline). */
getIssueTimeline(index: number): Promise<LifecycleEvent[]>
/** Every label defined on the repo (id + name), for name→id resolution. */
listLabels(): Promise<GiteaLabel[]>
/** Replace an issue's entire label set with the given label ids. Write. */
setIssueLabels(index: number, labelIds: number[]): Promise<void>
/** Open a new issue with a title, optional body, and label ids. Write. */
createIssue(input: { title: string; body?: string; labelIds?: number[] }): Promise<GiteaIssue>
/** Read a repo file's base64 content + blob sha; null if it (or the repo) is absent. */
getFile(path: string): Promise<{ contentBase64: string; sha: string } | null>
/** Create or update a repo file with base64 content (pass `sha` to update). Write. */
putFile(path: string, input: { contentBase64: string; message: string; sha?: string }): Promise<void>
}
/** Map raw gitea issue JSON to the normalized domain shape. Pure. */
@@ -144,18 +155,24 @@ export function createGiteaClient(config: GiteaConfig, fetchImpl: FetchLike): Gi
const apiBase = `${config.baseUrl.replace(/\/+$/, '')}/api/v1`
const repoBase = `${apiBase}/repos/${config.owner}/${config.repo}`
async function request(path: string): Promise<unknown> {
async function request(path: string, init?: { method?: string; body?: unknown }): Promise<unknown> {
const method = init?.method ?? 'GET'
const hasBody = init?.body !== undefined
const res = await fetchImpl(`${repoBase}${path}`, {
method,
headers: {
Authorization: `token ${config.token}`,
Accept: 'application/json',
...(hasBody ? { 'Content-Type': 'application/json' } : {}),
},
body: hasBody ? JSON.stringify(init!.body) : undefined,
})
if (!res.ok) {
const body = await res.text().catch(() => '')
throw new GiteaApiError(res.status, `GET ${path} failed (${res.status})`, body)
throw new GiteaApiError(res.status, `${method} ${path} failed (${res.status})`, body)
}
return res.json()
// writes may reply 204 No Content
return res.status === 204 ? null : res.json()
}
/** Follow gitea's page-limit pagination until a short page is returned. */
@@ -200,5 +217,44 @@ export function createGiteaClient(config: GiteaConfig, fetchImpl: FetchLike): Gi
)
return normalizeTimeline(raw)
},
async listLabels() {
const raw = await requestAll<{ id: number; name: string }>(
(page) => `/labels?page=${page}&limit=${PAGE_LIMIT}`,
)
return raw.map((l) => ({ id: l.id, name: l.name }))
},
async setIssueLabels(index, labelIds) {
await request(`/issues/${index}/labels`, { method: 'PUT', body: { labels: labelIds } })
},
async createIssue(input) {
const raw = (await request('/issues', {
method: 'POST',
body: { title: input.title, body: input.body ?? '', labels: input.labelIds ?? [] },
})) as RawIssue
return normalizeIssue(raw)
},
async getFile(path) {
const res = await fetchImpl(`${repoBase}/contents/${path}`, {
headers: { Authorization: `token ${config.token}`, Accept: 'application/json' },
})
if (res.status === 404) return null
if (!res.ok) {
const body = await res.text().catch(() => '')
throw new GiteaApiError(res.status, `GET contents/${path} failed (${res.status})`, body)
}
const json = (await res.json()) as { content?: string; sha: string }
return { contentBase64: (json.content ?? '').replace(/\n/g, ''), sha: json.sha }
},
async putFile(path, input) {
await request(`/contents/${path}`, {
method: input.sha ? 'PUT' : 'POST',
body: { content: input.contentBase64, message: input.message, ...(input.sha ? { sha: input.sha } : {}) },
})
},
}
}

View File

@@ -35,6 +35,12 @@ export interface GiteaHttpResponse {
export type FetchLike = (url: string, init?: GiteaRequestInit) => Promise<GiteaHttpResponse>
/** A repo label — just the id + name we need for name→id resolution. */
export interface GiteaLabel {
id: number
name: string
}
/** Milestone as referenced from an issue (not the full milestone resource). */
export interface GiteaMilestoneRef {
id: number

View File

@@ -17,11 +17,15 @@ export type {
GiteaConfig,
GiteaHttpResponse,
GiteaIssue,
GiteaLabel,
GiteaMilestone,
GiteaMilestoneRef,
GiteaRequestInit,
} from './gitea/types.js'
export { describeChange, planIssueChange, proposalsFor } from './changes/apply-changes-v0.js'
export type { ChangeProposal, IssueChange, LabelPlan, ProposeChangeArgs } from './changes/apply-changes-v0.js'
export {
inferColumnV0,
inferLifecycle,
@@ -81,6 +85,30 @@ export { pickModel } from './agent/model-router.js'
export type { ModelRouter, TaskKind } from './agent/model-router.js'
export { runAgentTurn } from './agent/agent-loop.js'
export type { AgentStep, AgentTurn, ToolExecutor } from './agent/agent-loop.js'
export { QUERY_PROJECT_TOOL, REGINALD_SYSTEM, REGINALD_TOOLS } from './agent/agent-tools.js'
export {
PROPOSE_CHANGE_TOOL,
QUERY_PROJECT_TOOL,
RECORD_DIRECTIVE_TOOL,
REGINALD_SYSTEM,
REGINALD_TOOLS,
} from './agent/agent-tools.js'
export { buildProjectView } from './agent/query-project.js'
export type { ProjectSnapshot, ProjectView, QueryFilters } from './agent/query-project.js'
export { CAPTURE_SYSTEM, captureWork, parseCaptureArgs, PROPOSE_ISSUES_TOOL } from './agent/capture-work.js'
export type { CaptureProposal, ProposedIssue } from './agent/capture-work.js'
export {
appendDirective,
makeDirectiveEntry,
parseDirectiveLog,
serializeDirective,
toDirectiveInput,
} from './directives/record-directive-v0.js'
export type {
DirectiveEntry,
DirectiveInput,
DirectiveKind,
DirectiveRecord,
DirectiveStatus,
DirectiveTarget,
} from './directives/record-directive-v0.js'