kiro-spec-engine 1.47.1 → 1.47.3

package/CHANGELOG.md

@@ -8,6 +8,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [Unreleased]
 
 ### Added
+- **Controller checkpoint resume**: `close-loop-controller` now supports `--controller-resume <latest|id|file>` plus persisted controller session snapshots (`.kiro/auto/close-loop-controller-sessions`) with retention controls (`--controller-session-id`, `--controller-session-keep`, `--controller-session-older-than-days`, `--no-controller-session`).
+- **Controller concurrency lease lock**: Added queue lease lock controls (`--controller-lock-file`, `--controller-lock-ttl-seconds`, `--no-controller-lock`) to prevent concurrent queue corruption and allow stale-lock takeover under bounded TTL.
+- **Persistent close-loop queue controller**: Added `kse auto close-loop-controller [queue-file]` with queue drain/poll runtime (`--queue-format`, `--dequeue-limit`, `--wait-on-empty`, `--poll-seconds`, runtime budgets, done/failed archives) so broad goals can be continuously executed through `close-loop-program` without manual re-invocation.
 - **KPI sample generator command**: Added `kse value metrics sample` to generate a ready-to-use KPI input JSON scaffold (`kpi-input.json`) for first-time observability runs.
 - **Value observability documentation track**: Added dedicated EN/ZH guides for KPI snapshot/baseline/trend workflow (`docs/value-observability-guide.md`, `docs/zh/value-observability-guide.md`) and wired entry links from README + docs indexes for faster discovery of measurable delivery capabilities.
 - **Release communication assets**: Added bilingual v1.46.2 release notes (`docs/releases/v1.46.2.md`, `docs/zh/releases/v1.46.2.md`) and a reusable pre-release checklist (`docs/release-checklist.md`).
@@ -88,6 +91,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - **Live orchestration status streaming**: Added event/interval-driven status persistence callback support in `runOrchestration()` and wired `auto close-loop` live progress output (`--no-stream` to disable).
 
 ### Changed
+- **Controller queue hygiene default**: `close-loop-controller` now deduplicates duplicate broad goals by default (`--no-controller-dedupe` to preserve raw duplicates), and summary telemetry includes dedupe/lock/session metadata.
 - **Positioning and onboarding messaging**: Strengthened EN/ZH README and quick-start docs with explicit kse advantage matrix, 90-second value proof, and KPI observability positioning to improve first-contact clarity.
 - **CLI first-screen positioning text**: Updated `kse --help` top description in EN/ZH locales to reflect current core strengths: Spec workflow, orchestration, and KPI observability.
 - **Offline onboarding consistency**: Refreshed `START_HERE.txt`, `INSTALL_OFFLINE.txt`, and `docs/OFFLINE_INSTALL.md` to v1.46.2 guidance and aligned quick-start prerequisites with current runtime requirement (Node.js >= 16).
@@ -100,6 +104,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - **Autonomous operator UX**: Expanded docs with semantic decomposition and live stream behavior for close-loop command usage.
 
 ### Fixed
+- **Controller summary semantics**: `close-loop-controller` now reports final `pending_goals` from the persisted queue snapshot and only marks `cycle-limit-reached` as exhausted when pending work remains (or empty-polling mode explicitly consumes cycle budget).
 - **npm package hygiene**: Excluded transient Python bytecode artifacts (`__pycache__`, `*.pyc/pyo/pyd`) from published package contents to reduce package noise and size.
 - **Documentation contact placeholders**: Replaced `yourusername` placeholder repository links in onboarding docs with the canonical project URL and removed stale example email contact.
 - **Jest force-exit dependency**: Removed `forceExit` from `jest.config.js` and `jest.config.ci.js`; test scripts now complete without explicit force-exit configuration.

package/README.md

@@ -57,6 +57,7 @@ graph LR
 | Multi-agent scale-out | DAG orchestration (`orchestrate run/status/stop`) | Parallel delivery without manual terminal juggling |
 | Auto portfolio decomposition | Master/Sub spec split + dependency wiring | Complex goals become executable parallel plans automatically |
 | Program self-healing closure | `auto close-loop-program` auto recovery + learned remediation memory | Failed/partial runs continue autonomously until bounded convergence |
+| Queue-driven autonomous runtime | `auto close-loop-controller` persistent queue drain | Program-scale backlogs continue without manual re-triggering |
 | Autonomous convergence governance | Program gate (`min-success-rate` + `max-risk-level`) + audit JSON output | Program completion is policy-verified and traceable |
 | Measurable outcomes | KPI automation (`value metrics snapshot/baseline/trend`) | Delivery quality can be tracked week-over-week |
 | Tool-agnostic adoption | Works across Claude/Cursor/Windsurf/Copilot/Kiro | No lock-in to a single AI IDE |
@@ -532,6 +533,8 @@ kse create-spec <name>             # Legacy: create empty Spec folder only
 kse auto close-loop "<goal>"       # Auto split goal into master/sub specs and execute to closure
 kse auto close-loop "<goal>" --dry-run --json  # Preview decomposition plan only
 kse auto close-loop-program "<goal>" --program-govern-until-stable --program-govern-use-action 1 --json # Program-level recovery + governance with remediation action execution
+kse auto close-loop-controller .kiro/auto/program-queue.lines --wait-on-empty --dequeue-limit 2 --json # Queue-driven autonomous controller for broad-goal backlogs
+kse auto close-loop-controller --controller-resume latest --json # Resume autonomous controller from latest persisted checkpoint
 
 # Spec workflow (recommended)
 kse spec bootstrap --name <spec> --non-interactive         # Generate requirements/design/tasks draft
@@ -736,6 +739,6 @@ A deep conversation about AI development trends, Neo-Confucian philosophy, and s
 
 ---
 
-**Version**: 1.46.2  
-**Last Updated**: 2026-02-14
+**Version**: 1.47.3  
+**Last Updated**: 2026-02-15
 

package/README.zh.md

@@ -55,6 +55,7 @@ graph LR
 | 多 Agent 扩展 | DAG 编排（`orchestrate run/status/stop`） | 并行交付，不再手工开多终端 |
 | 自动主从拆分 | 目标自动拆成 Master/Sub Spec 并建立依赖 | 复杂需求自动转成可并行执行计划 |
 | 程序级自愈闭环 | `auto close-loop-program` 自动恢复 + 策略记忆 | 失败/部分失败会自动续跑，直到有界收敛 |
+| 队列驱动自治执行 | `auto close-loop-controller` 持续消费目标队列 | 程序级积压目标无需人工反复触发 |
 | 自治收敛治理 | Program Gate（最小成功率 + 最大风险）+ 审计 JSON 导出 | 完成结果可按策略验收且全程可追溯 |
 | 结果可量化 | KPI 自动化（`value metrics snapshot/baseline/trend`） | 周度趋势可追踪、可审计 |
 | 工具无锁定 | 兼容 Claude/Cursor/Windsurf/Copilot/Kiro | 团队可保留现有 AI 工具链 |
@@ -449,6 +450,8 @@ kse create-spec <name>             # 兼容旧版：仅创建空 Spec 目录
 kse auto close-loop "<目标>"        # 自动拆分 Master/Sub Spec 并推进到完成态
 kse auto close-loop "<目标>" --dry-run --json  # 仅预览拆分与依赖计划
 kse auto close-loop-program "<目标>" --program-govern-until-stable --program-govern-use-action 1 --json # 程序级自动恢复 + 治理循环（含 remediation action 执行）直到稳定
+kse auto close-loop-controller .kiro/auto/program-queue.lines --wait-on-empty --dequeue-limit 2 --json # 队列驱动自治控制器，持续处理广义目标积压
+kse auto close-loop-controller --controller-resume latest --json # 从最近一次 controller 检查点恢复自治推进
 
 # Spec 工作流（推荐）
 kse spec bootstrap --name <spec> --non-interactive          # 生成 requirements/design/tasks 初稿
@@ -600,7 +603,7 @@ kse spec bootstrap --name 01-00-my-first-feature --non-interactive
 
 ---
 
-**版本**：1.46.2  
-**最后更新**：2026-02-14
+**版本**：1.47.3  
+**最后更新**：2026-02-15
 
 

package/docs/autonomous-control-guide.md

@@ -87,6 +87,24 @@ kse auto close-loop-program \
   --program-audit-out .kiro/reports/close-loop-program-audit.json \
   --json
 
+# Controller command: drain queue goals with autonomous close-loop-program runtime
+kse auto close-loop-controller .kiro/auto/program-queue.lines \
+  --dequeue-limit 2 \
+  --max-cycles 20 \
+  --controller-done-file .kiro/auto/program-done.lines \
+  --controller-failed-file .kiro/auto/program-failed.lines \
+  --json
+
+# Persistent controller mode: keep polling queue and execute new goals automatically
+kse auto close-loop-controller .kiro/auto/program-queue.lines \
+  --wait-on-empty \
+  --poll-seconds 30 \
+  --max-cycles 1000 \
+  --max-minutes 240
+
+# Resume from latest persisted controller session
+kse auto close-loop-controller --controller-resume latest --json
+
 # Recovery command: consume diagnostics and auto-recover unresolved goals
 kse auto close-loop-recover latest --json
 kse auto close-loop-recover .kiro/auto/close-loop-batch-summaries/batch-20260215090000.json \
@@ -282,6 +300,23 @@ Close-loop program command:
 - Summary also includes `program_diagnostics` (`failure_clusters`, `remediation_actions`) to drive automatic follow-up and convergence recovery.
 - Summary includes `program_coordination` (master/sub topology, unresolved goal indexes, scheduler telemetry) and `auto_recovery` metadata.
 
+Close-loop controller command:
+- Command: `kse auto close-loop-controller [queue-file]`
+- Runs a queue-driven autonomous loop: dequeue broad goals, execute each by `close-loop-program`, persist remaining queue, and continue until stop condition.
+- Queue file defaults to `.kiro/auto/close-loop-controller-goals.lines`; supports `auto|json|lines` parsing via `--queue-format`.
+- `--controller-resume <session-or-file>` resumes queue/controller context from persisted controller session (`latest`, session id, or file path).
+- Duplicate broad goals are deduped by default; use `--no-controller-dedupe` to preserve raw queue duplicates.
+- `--dequeue-limit <n>` controls how many queued goals are consumed in one cycle (`1-100`, default `1`).
+- `--wait-on-empty` + `--poll-seconds <n>` enables long-running poll mode for continuously appended program queues.
+- `--max-cycles <n>` + `--max-minutes <n>` bound controller runtime to prevent unbounded loops.
+- Controller lease lock is enabled by default to prevent concurrent queue corruption (`--controller-lock-file`, `--controller-lock-ttl-seconds`, `--no-controller-lock`).
+- `--stop-on-goal-failure` stops controller immediately on the first failed dequeued goal.
+- `--controller-done-file` / `--controller-failed-file` append per-goal archive lines for downstream ops and replay.
+- Controller summaries are persisted by default (`.kiro/auto/close-loop-controller-sessions`); use `--controller-session-id`, `--controller-session-keep`, `--controller-session-older-than-days`, `--no-controller-session` as needed.
+- `--controller-print-program-summary` prints each nested program summary in realtime during controller execution.
+- Supports full program governance/gate/recovery policy controls (`--program-*`, `--batch-*`, `--continue-on-error`, `--recovery-memory-scope`, `--dry-run`, `--json`).
+- Output summary includes `history`, `results`, final `pending_goals`, `stop_reason`, `exhausted`, dedupe/lock/session, and resume-source telemetry.
+
 Close-loop recovery command:
 - Command: `kse auto close-loop-recover [summary]` (defaults to `latest` if summary is omitted).
 - Loads unresolved goals from a prior summary and applies selected remediation strategy patch automatically.

package/docs/command-reference.md

@@ -2,7 +2,7 @@
 
 > Quick reference for all kse commands
 
-**Version**: 1.46.2  
+**Version**: 1.47.3  
 **Last Updated**: 2026-02-15
 
 ---
@@ -344,6 +344,24 @@ kse auto close-loop-program \
   --program-audit-out .kiro/reports/close-loop-program-audit.json \
   --json
 
+# Controller command: drain queued broad goals with close-loop-program runtime
+kse auto close-loop-controller .kiro/auto/program-queue.lines \
+  --dequeue-limit 2 \
+  --max-cycles 20 \
+  --controller-done-file .kiro/auto/program-done.lines \
+  --controller-failed-file .kiro/auto/program-failed.lines \
+  --json
+
+# Persistent controller mode: keep polling queue and execute new goals automatically
+kse auto close-loop-controller .kiro/auto/program-queue.lines \
+  --wait-on-empty \
+  --poll-seconds 30 \
+  --max-cycles 1000 \
+  --max-minutes 240
+
+# Resume from latest persisted controller session
+kse auto close-loop-controller --controller-resume latest --json
+
 # Recovery command: replay unresolved goals from summary using remediation action strategy
 kse auto close-loop-recover latest --json
 kse auto close-loop-recover .kiro/auto/close-loop-batch-summaries/batch-20260215090000.json \
@@ -548,6 +566,29 @@ Close-loop program (`kse auto close-loop-program "<goal>"`) options:
 - Program summary includes `program_diagnostics` with `failure_clusters` and `remediation_actions` (prioritized follow-up commands for convergence).
 - Program summary includes `program_coordination` (master/sub topology, unresolved goal indexes, scheduler snapshot) and `auto_recovery` metadata.
 
+Close-loop controller (`kse auto close-loop-controller [queue-file]`) options:
+- `queue-file`: optional queue file path (default `.kiro/auto/close-loop-controller-goals.lines`)
+- `--controller-resume <session-or-file>`: resume from persisted controller session (`latest`, session id, or file path)
+- `--queue-format <auto|json|lines>`: queue parser mode (default `auto`)
+- `--no-controller-dedupe`: disable duplicate broad-goal deduplication (default dedupe enabled)
+- `--dequeue-limit <n>`: consume up to `n` goals per controller cycle (`1-100`, default `1`)
+- `--wait-on-empty`: keep polling when queue is empty instead of stopping
+- `--poll-seconds <n>`: polling interval for `--wait-on-empty` (`1-3600`, default `30`)
+- `--max-cycles <n>`: max controller cycles (`1-100000`, default `1000`)
+- `--max-minutes <n>`: elapsed-time budget in minutes (`1-10080`, default `120`)
+- `--controller-lock-file <path>`: explicit lease lock file (default `<queue-file>.lock`)
+- `--controller-lock-ttl-seconds <n>`: stale lock takeover threshold (`10-86400`, default `1800`)
+- `--no-controller-lock`: disable controller lease lock (unsafe for concurrent controllers)
+- `--stop-on-goal-failure`: stop immediately when one dequeued goal fails
+- `--controller-session-id <id>`: set explicit persisted controller session id
+- `--controller-session-keep <n>` / `--controller-session-older-than-days <n>`: retention policy for persisted controller sessions
+- `--no-controller-session`: disable controller session persistence
+- `--controller-out <path>`: write controller summary JSON
+- `--controller-done-file <path>` / `--controller-failed-file <path>`: append completed/failed goals into line archives
+- `--controller-print-program-summary`: print each nested `close-loop-program` summary during controller execution
+- Supports program execution controls (`--program-*`, `--batch-*`, `--continue-on-error`, `--recovery-memory-scope`, `--dry-run`, `--json`) and runs each dequeued queue goal through full autonomous program flow.
+- Summary includes controller telemetry (`history`, `results`, final `pending_goals`, `stop_reason`, `exhausted`, dedupe/lock/session metadata) plus optional done/failed archive file paths.
+
 Close-loop recovery (`kse auto close-loop-recover [summary]`) options:
 - `summary`: optional summary file path; defaults to `latest` persisted batch summary
 - `--use-action <n>`: choose remediation action index from diagnostics (`1` by default)