peaks-cli 1.2.6 → 1.2.8

package/dist/src/services/progress/progress-service.d.ts

@@ -0,0 +1,179 @@
+/**
+ * Sub-agent progress surfacing for the RD/QA sub-agents in
+ * `peaks-solo`'s Swarm phase. A sub-agent (or the LLM via the
+ * `peaks progress step` CLI) writes a stable JSON file at
+ * `.peaks/<sid>/system/subagent-progress.json`. The user-side
+ * `peaks progress watch` CLI polls this file in a separate
+ * terminal tab and renders elapsed / spinner / sub-step. The
+ * `peaks progress start` CLI auto-spawns the watch in a new
+ * terminal window so the user does not have to remember to do
+ * it.
+ *
+ * Token cost design (the binding constraint of this feature):
+ *   - The LLM side (step CLI) writes the file at most once per
+ *     phase transition. That is approximately one Bash call per
+ *     RD/QA sub-step. In a typical 5-step sub-agent slice the
+ *     cost is < 10 output tokens.
+ *   - The watch side polls a local file, not the LLM. Zero token
+ *     cost.
+ *   - The auto-spawn side (start CLI) is invoked once per
+ *     session by the LLM at the first phase transition. One
+ *     Bash call. The user closes the new terminal at any time;
+ *     no further side effects.
+ *
+ * Net: the LLM pays a one-time < 10 token cost per slice to
+ * give the user real-time progress visibility. The user pays
+ * zero manual setup.
+ *
+ * This module is pure filesystem. It does NOT import the LLM
+ * harness, does NOT spawn terminals, and does NOT talk to any
+ * IPC. Those are concerns of the CLI layer (../cli/commands/
+ * progress-commands.ts and hooks-settings-service.ts).
+ */
+export type SubAgentProgressPhase = 'starting' | 'running' | 'verifying' | 'completing' | 'finished' | 'failed' | 'idle';
+export type SubAgentProgressStep = {
+    /** ISO-8601 timestamp at which the sub-agent started. */
+    startedAt: string;
+    /** ISO-8601 timestamp at which the sub-agent finished (or now, if still running). */
+    updatedAt: string;
+    /** Free-form human-readable sub-step label, e.g. "running test/ut". */
+    step: string;
+    /** Current phase bucket. `idle` is the pre-start sentinel. */
+    phase: SubAgentProgressPhase;
+    /** When set, the sub-agent is finished and reports the verdict here. */
+    verdict?: 'pass' | 'return-to-rd' | 'blocked';
+    /** Optional count of in-scope files touched, assertions run, etc. */
+    counts?: {
+        filesTouched?: number;
+        testsRun?: number;
+    };
+};
+export type SubAgentProgress = {
+    version: 1;
+    sessionId: string;
+    outerSessionId?: string;
+    /** Outer-agent role that owns the slice (e.g. "rd", "qa"). */
+    role: string;
+    /** Per-slice identifier. */
+    requestId: string;
+    /** When the sub-agent entered the first non-idle state. */
+    startedAt: string;
+    /** When the sub-agent last touched the file. */
+    updatedAt: string;
+    /** Current step. */
+    current: SubAgentProgressStep;
+    /**
+     * History of completed steps. Length is unbounded; the watch
+     * tool renders the most recent N. Kept for after-the-fact
+     * forensics ("how long did step 3 take?").
+     */
+    history: SubAgentProgressStep[];
+};
+export type ReadProgressOptions = {
+    projectRoot: string;
+};
+export type ReadProgressResult = {
+    ok: true;
+    data: SubAgentProgress;
+    path: string;
+} | {
+    ok: false;
+    reason: 'no-binding' | 'no-progress-file' | 'invalid-json';
+};
+export type WriteProgressOptions = {
+    projectRoot: string;
+    requestId: string;
+    role: string;
+    step: string;
+    phase: SubAgentProgressPhase;
+    verdict?: 'pass' | 'return-to-rd' | 'blocked';
+    counts?: SubAgentProgressStep['counts'];
+    outerSessionId?: string;
+};
+/**
+ * Read the current progress file. Returns a tagged result so
+ * the CLI can map each failure mode to a distinct nextActions
+ * hint (no-binding → run peaks workspace init; no-progress-file
+ * → sub-agent has not started yet; invalid-json → the LLM wrote
+ * garbage; recover by writing a fresh file).
+ */
+export declare function readSubAgentProgress(options: ReadProgressOptions): ReadProgressResult;
+/**
+ * Append-or-replace the current step. Idempotent on identical
+ * (step, phase) — the file is rewritten with the same payload
+ * (no new history entry) so heartbeats from the same phase do
+ * not pollute the history. Phase transitions append a new step
+ * to the history and replace `current`.
+ */
+export declare function writeSubAgentProgress(options: WriteProgressOptions): SubAgentProgress;
+/**
+ * Resolve the project root for a CLI invocation: --project
+ * override wins, otherwise the canonical git-root promotion
+ * (so the sub-agent's writes land in the same `.peaks/<sid>/`
+ * the user's manual CLI would use). Re-exports the same
+ * helper peaks workspace init / session rotate already use
+ * for symmetry.
+ */
+export declare function resolveProgressProjectRoot(override: string | undefined, cwd: string): string;
+/**
+ * Compute the absolute path to the progress file for a given
+ * project root, for callers that need to display / fs.watch it
+ * (the watch banner, the auto-spawn helper, the close command).
+ * This MUST agree with `progressPath` — the read/write helpers
+ * resolve through `progressPath` and use the session sub-directory,
+ * so the displayed path does too. Without this agreement the
+ * watch banner would point at a path the file is never written
+ * to, and the user would `cat` an empty file.
+ */
+export declare function subAgentProgressPath(projectRoot: string): string;
+/**
+ * Compute the absolute path to the spawn record for a given
+ * project root. Exported so `peaks progress close` (and the
+ * start command's success payload) can advertise the on-disk
+ * location without re-deriving the session sub-directory.
+ */
+export declare function subAgentSpawnPath(projectRoot: string): string;
+export type ProgressSpawnRecord = {
+    version: 1;
+    sessionId: string;
+    pid: number;
+    platform: NodeJS.Platform;
+    command: string;
+    args: string[];
+    spawnedAt: string;
+    reason?: string;
+    /** The title we asked the terminal emulator to set. */
+    windowTitle: string;
+};
+export type WriteSpawnRecordOptions = {
+    projectRoot: string;
+    pid: number;
+    platform: NodeJS.Platform;
+    command: string;
+    args: string[];
+    reason?: string;
+    windowTitle: string;
+};
+export declare function writeSpawnRecord(options: WriteSpawnRecordOptions): ProgressSpawnRecord | null;
+export type ReadSpawnRecordResult = {
+    ok: true;
+    data: ProgressSpawnRecord;
+    path: string;
+} | {
+    ok: false;
+    reason: 'no-binding' | 'no-spawn-record' | 'invalid-json';
+};
+export declare function readSpawnRecord(projectRoot: string): ReadSpawnRecordResult;
+export declare function clearSpawnRecord(projectRoot: string): boolean;
+export type PhaseClosingTrigger = 'finished' | 'failed';
+/**
+ * True if a transition into the given phase should auto-close
+ * the spawned watch window. `finished` and `failed` both
+ * indicate the sub-agent is done; a `blocked` verdict on a
+ * `finished` step is intentionally NOT a close trigger
+ * because a blocked slice usually means the user needs to
+ * read the watch output before deciding what to do. The CLI
+ * layer reads `data.current.phase`, not the verdict, so this
+ * helper is the only close-decision source of truth.
+ */
+export declare function phaseAutoClosesSpawn(phase: SubAgentProgressPhase): boolean;

package/dist/src/services/progress/progress-service.js

@@ -0,0 +1,276 @@
+/**
+ * Sub-agent progress surfacing for the RD/QA sub-agents in
+ * `peaks-solo`'s Swarm phase. A sub-agent (or the LLM via the
+ * `peaks progress step` CLI) writes a stable JSON file at
+ * `.peaks/<sid>/system/subagent-progress.json`. The user-side
+ * `peaks progress watch` CLI polls this file in a separate
+ * terminal tab and renders elapsed / spinner / sub-step. The
+ * `peaks progress start` CLI auto-spawns the watch in a new
+ * terminal window so the user does not have to remember to do
+ * it.
+ *
+ * Token cost design (the binding constraint of this feature):
+ *   - The LLM side (step CLI) writes the file at most once per
+ *     phase transition. That is approximately one Bash call per
+ *     RD/QA sub-step. In a typical 5-step sub-agent slice the
+ *     cost is < 10 output tokens.
+ *   - The watch side polls a local file, not the LLM. Zero token
+ *     cost.
+ *   - The auto-spawn side (start CLI) is invoked once per
+ *     session by the LLM at the first phase transition. One
+ *     Bash call. The user closes the new terminal at any time;
+ *     no further side effects.
+ *
+ * Net: the LLM pays a one-time < 10 token cost per slice to
+ * give the user real-time progress visibility. The user pays
+ * zero manual setup.
+ *
+ * This module is pure filesystem. It does NOT import the LLM
+ * harness, does NOT spawn terminals, and does NOT talk to any
+ * IPC. Those are concerns of the CLI layer (../cli/commands/
+ * progress-commands.ts and hooks-settings-service.ts).
+ */
+import { existsSync, mkdirSync, readFileSync, unlinkSync, writeFileSync } from 'node:fs';
+import { dirname, join, resolve } from 'node:path';
+import { getSessionIdCanonical } from '../session/session-manager.js';
+import { findProjectRoot } from '../config/config-safety.js';
+const PROGRESS_REL_PATH = 'system/subagent-progress.json';
+const SPAWN_REL_PATH = 'system/progress-spawn.json';
+function progressPath(projectRoot) {
+    // The progress file lives under the *session* directory, not
+    // directly under .peaks/. Every other per-slice artefact
+    // (rd/tech-doc.md, qa/test-cases/<rid>.md, prd/requests/<rid>.md,
+    // memory/, openspec/) lives under .peaks/<sid>/, so progress
+    // should too. Without the session prefix, a session rotation
+    // would orphan the file in the project root, and switching
+    // sessions would have the watch reading the wrong slice's
+    // progress.
+    const sessionId = getSessionIdCanonical(projectRoot);
+    const subDir = sessionId ?? 'unbound';
+    return join(projectRoot, '.peaks', subDir, PROGRESS_REL_PATH);
+}
+function ensureParentDir(path) {
+    const dir = dirname(path);
+    if (!existsSync(dir)) {
+        mkdirSync(dir, { recursive: true });
+    }
+}
+function nowIso() {
+    return new Date().toISOString();
+}
+/**
+ * Read the current progress file. Returns a tagged result so
+ * the CLI can map each failure mode to a distinct nextActions
+ * hint (no-binding → run peaks workspace init; no-progress-file
+ * → sub-agent has not started yet; invalid-json → the LLM wrote
+ * garbage; recover by writing a fresh file).
+ */
+export function readSubAgentProgress(options) {
+    const sessionId = getSessionIdCanonical(options.projectRoot);
+    if (sessionId === null) {
+        return { ok: false, reason: 'no-binding' };
+    }
+    const path = progressPath(options.projectRoot);
+    if (!existsSync(path)) {
+        return { ok: false, reason: 'no-progress-file' };
+    }
+    try {
+        const data = JSON.parse(readFileSync(path, 'utf8'));
+        if (data.version !== 1 || typeof data.sessionId !== 'string') {
+            return { ok: false, reason: 'invalid-json' };
+        }
+        return { ok: true, data, path };
+    }
+    catch {
+        return { ok: false, reason: 'invalid-json' };
+    }
+}
+/**
+ * Append-or-replace the current step. Idempotent on identical
+ * (step, phase) — the file is rewritten with the same payload
+ * (no new history entry) so heartbeats from the same phase do
+ * not pollute the history. Phase transitions append a new step
+ * to the history and replace `current`.
+ */
+export function writeSubAgentProgress(options) {
+    const existing = readSubAgentProgress({ projectRoot: options.projectRoot });
+    const now = nowIso();
+    const path = progressPath(options.projectRoot);
+    if (existing.ok) {
+        const prev = existing.data;
+        // Heartbeat on the same current step: just bump updatedAt, do
+        // NOT add a history entry. The shape of `current` is preserved.
+        if (prev.current.step === options.step && prev.current.phase === options.phase) {
+            const next = {
+                ...prev,
+                updatedAt: now,
+                current: {
+                    ...prev.current,
+                    updatedAt: now,
+                    ...(options.verdict !== undefined ? { verdict: options.verdict } : {}),
+                    ...(options.counts !== undefined ? { counts: options.counts } : {})
+                }
+            };
+            ensureParentDir(path);
+            writeFileSync(path, JSON.stringify(next, null, 2) + '\n', 'utf8');
+            return next;
+        }
+        // Real phase / step transition: archive the prior current
+        // into history, install the new current.
+        const archived = {
+            ...prev.current,
+            updatedAt: now
+        };
+        const next = {
+            ...prev,
+            updatedAt: now,
+            current: {
+                startedAt: now,
+                updatedAt: now,
+                step: options.step,
+                phase: options.phase,
+                ...(options.verdict !== undefined ? { verdict: options.verdict } : {}),
+                ...(options.counts !== undefined ? { counts: options.counts } : {})
+            },
+            history: [...prev.history, archived]
+        };
+        ensureParentDir(path);
+        writeFileSync(path, JSON.stringify(next, null, 2) + '\n', 'utf8');
+        return next;
+    }
+    // No prior file: this is the first write. Bootstrap a fresh
+    // progress doc. The sessionId is whatever the binding points
+    // at, so cross-session confusion is impossible.
+    const sessionId = getSessionIdCanonical(options.projectRoot) ?? 'unbound';
+    const fresh = {
+        version: 1,
+        sessionId,
+        ...(options.outerSessionId !== undefined ? { outerSessionId: options.outerSessionId } : {}),
+        role: options.role,
+        requestId: options.requestId,
+        startedAt: now,
+        updatedAt: now,
+        current: {
+            startedAt: now,
+            updatedAt: now,
+            step: options.step,
+            phase: options.phase,
+            ...(options.verdict !== undefined ? { verdict: options.verdict } : {}),
+            ...(options.counts !== undefined ? { counts: options.counts } : {})
+        },
+        history: []
+    };
+    ensureParentDir(path);
+    writeFileSync(path, JSON.stringify(fresh, null, 2) + '\n', 'utf8');
+    return fresh;
+}
+/**
+ * Resolve the project root for a CLI invocation: --project
+ * override wins, otherwise the canonical git-root promotion
+ * (so the sub-agent's writes land in the same `.peaks/<sid>/`
+ * the user's manual CLI would use). Re-exports the same
+ * helper peaks workspace init / session rotate already use
+ * for symmetry.
+ */
+export function resolveProgressProjectRoot(override, cwd) {
+    if (override !== undefined)
+        return override;
+    return findProjectRoot(cwd) ?? cwd;
+}
+/**
+ * Compute the absolute path to the progress file for a given
+ * project root, for callers that need to display / fs.watch it
+ * (the watch banner, the auto-spawn helper, the close command).
+ * This MUST agree with `progressPath` — the read/write helpers
+ * resolve through `progressPath` and use the session sub-directory,
+ * so the displayed path does too. Without this agreement the
+ * watch banner would point at a path the file is never written
+ * to, and the user would `cat` an empty file.
+ */
+export function subAgentProgressPath(projectRoot) {
+    return progressPath(resolve(projectRoot));
+}
+/**
+ * Compute the absolute path to the spawn record for a given
+ * project root. Exported so `peaks progress close` (and the
+ * start command's success payload) can advertise the on-disk
+ * location without re-deriving the session sub-directory.
+ */
+export function subAgentSpawnPath(projectRoot) {
+    const sessionId = getSessionIdCanonical(projectRoot);
+    const subDir = sessionId ?? 'unbound';
+    return join(projectRoot, '.peaks', subDir, SPAWN_REL_PATH);
+}
+function spawnRecordPath(projectRoot) {
+    const sessionId = getSessionIdCanonical(projectRoot);
+    const subDir = sessionId ?? 'unbound';
+    return join(projectRoot, '.peaks', subDir, SPAWN_REL_PATH);
+}
+export function writeSpawnRecord(options) {
+    const sessionId = getSessionIdCanonical(options.projectRoot);
+    if (sessionId === null)
+        return null;
+    const now = nowIso();
+    const record = {
+        version: 1,
+        sessionId,
+        pid: options.pid,
+        platform: options.platform,
+        command: options.command,
+        args: options.args,
+        spawnedAt: now,
+        ...(options.reason !== undefined ? { reason: options.reason } : {}),
+        windowTitle: options.windowTitle
+    };
+    const path = spawnRecordPath(options.projectRoot);
+    ensureParentDir(path);
+    writeFileSync(path, JSON.stringify(record, null, 2) + '\n', 'utf8');
+    return record;
+}
+export function readSpawnRecord(projectRoot) {
+    const sessionId = getSessionIdCanonical(projectRoot);
+    if (sessionId === null)
+        return { ok: false, reason: 'no-binding' };
+    const path = spawnRecordPath(projectRoot);
+    if (!existsSync(path))
+        return { ok: false, reason: 'no-spawn-record' };
+    try {
+        const data = JSON.parse(readFileSync(path, 'utf8'));
+        if (data.version !== 1 || typeof data.pid !== 'number') {
+            return { ok: false, reason: 'invalid-json' };
+        }
+        return { ok: true, data, path };
+    }
+    catch {
+        return { ok: false, reason: 'invalid-json' };
+    }
+}
+export function clearSpawnRecord(projectRoot) {
+    const path = spawnRecordPath(projectRoot);
+    if (!existsSync(path))
+        return false;
+    try {
+        unlinkSync(path);
+        return true;
+    }
+    catch {
+        return false;
+    }
+}
+const PHASES_THAT_AUTO_CLOSE = new Set([
+    'finished',
+    'failed'
+]);
+/**
+ * True if a transition into the given phase should auto-close
+ * the spawned watch window. `finished` and `failed` both
+ * indicate the sub-agent is done; a `blocked` verdict on a
+ * `finished` step is intentionally NOT a close trigger
+ * because a blocked slice usually means the user needs to
+ * read the watch output before deciding what to do. The CLI
+ * layer reads `data.current.phase`, not the verdict, so this
+ * helper is the only close-decision source of truth.
+ */
+export function phaseAutoClosesSpawn(phase) {
+    return PHASES_THAT_AUTO_CLOSE.has(phase);
+}

package/dist/src/services/session/index.d.ts

@@ -1 +1 @@
-export { ensureSession, getSessionId, getCurrentSessionDir, listSessions, getSessionMeta, setSessionMeta, setSessionTitle, listSessionMetas, getProjectScanPath, hasProjectScan, setCurrentSessionBinding, type SessionInfo, type SessionMeta } from './session-manager.js';
+export { ensureSession, getSessionId, getCurrentSessionDir, listSessions, getSessionMeta, setSessionMeta, setSessionTitle, listSessionMetas, getProjectScanPath, hasProjectScan, setCurrentSessionBinding, rotateSessionBinding, type SessionInfo, type SessionMeta } from './session-manager.js';

package/dist/src/services/session/index.js

@@ -1 +1 @@
-export { ensureSession, getSessionId, getCurrentSessionDir, listSessions, getSessionMeta, setSessionMeta, setSessionTitle, listSessionMetas, getProjectScanPath, hasProjectScan, setCurrentSessionBinding } from './session-manager.js';
+export { ensureSession, getSessionId, getCurrentSessionDir, listSessions, getSessionMeta, setSessionMeta, setSessionTitle, listSessionMetas, getProjectScanPath, hasProjectScan, setCurrentSessionBinding, rotateSessionBinding } from './session-manager.js';

package/dist/src/services/session/session-manager.d.ts

@@ -19,7 +19,32 @@ export type SessionMeta = {
     createdAt: string;
     lastActivity?: string;
     projectRoot: string;
+    /**
+     * The outer (harness / IDE / plugin) session id that
+     * `ensureSession` was called from. Sourced from
+     * `PEAKS_OUTER_SESSION_ID` env var, with `CLAUDE_CODE_SESSION_ID`
+     * as a Claude-Code fallback. Stamped once at session creation;
+     * later presence writes can compare against this to detect an
+     * outer-session swap and AskUserQuestion the user about rolling
+     * a new peaks session. Sessions predating the field simply
+     * have it undefined; presence-mismatch detection skips those
+     * (no false positives on legacy data).
+     */
+    outerSessionId?: string;
 };
+/**
+ * Drop the project-level session binding (`.peaks/.session.json`)
+ * so the next `ensureSession()` call auto-generates a fresh
+ * session id. The on-disk session directory is left intact —
+ * rotating does NOT delete the user's data, it just unbinds the
+ * project from that session.
+ *
+ * Returns the id of the session that was unbound, or `null` if
+ * no binding was present. The caller is expected to do something
+ * with that — at minimum surface it in the CLI response so the
+ * user can find the directory again if they need to.
+ */
+export declare function rotateSessionBinding(projectRoot: string): string | null;
 /**
  * Bind the project's current session to the given session id by writing
  * `.peaks/.session.json`. The single-session binding is the source of truth
@@ -50,14 +75,6 @@ export declare function setSessionTitle(projectRoot: string, sessionId: string,
  * Returns sessions sorted by sessionId descending (most recent first).
  */
 export declare function listSessionMetas(projectRoot: string): SessionMeta[];
-/**
- * Get or create the current session for a project.
- * If a valid session already exists, returns it.
- * Otherwise, creates a new session with auto-generated ID.
- *
- * @param projectRoot - Root directory of the project
- * @returns Session ID (e.g., "2026-05-26-session-a3f8b1")
- */
 export declare function ensureSession(projectRoot: string): Promise<string>;
 /**
  * Get the current session ID without creating a new one.
@@ -67,6 +84,34 @@ export declare function ensureSession(projectRoot: string): Promise<string>;
  * @returns Session ID or null
  */
 export declare function getSessionId(projectRoot: string): string | null;
+/**
+ * Resolve the current session id with canonicalize-on-read
+ * semantics. This is the variant the progress subcommands
+ * (step / watch / start / close) use, because the legacy
+ * `getSessionId` returns null any time the stored
+ * `projectRoot` form differs from the caller-passed form
+ * (e.g. stored is "." from inside the project dir; caller
+ * is the absolute realpath). When `getSessionId` returns
+ * null, callers like `ensureSession` create a brand-new
+ * session and overwrite the binding — which is what the
+ * user observed as the "mid-dogfood rebind" bug.
+ *
+ * The fix is to canonicalize both sides of the compare
+ * (realpath, then optionally resolve relative stored
+ * against the caller's project root). The two forms of
+ * the same physical directory now compare equal, and the
+ * existing binding is found instead of being overwritten.
+ *
+ * Use this instead of `getSessionId` only when the
+ * caller is operating on a user-supplied `--project` flag
+ * and the binding may have been written by a CLI invocation
+ * that was running from inside the project dir (the common
+ * peaks-solo / peaks-sop scenario). Other modules depend
+ * on the strict-equality semantics of `getSessionId` (the
+ * "no binding" fallback path is part of their contract),
+ * so this variant is opt-in.
+ */
+export declare function getSessionIdCanonical(projectRoot: string): string | null;
 /**
  * Get the absolute path to the current session directory.
  * Creates the session if it doesn't exist.