Skip to content

Codex/canonical runtime deploy - #35

Open
lisihao wants to merge 780 commits into
mainfrom
codex/canonical-runtime-deploy
Open

Codex/canonical runtime deploy#35
lisihao wants to merge 780 commits into
mainfrom
codex/canonical-runtime-deploy

Conversation

@lisihao

@lisihao lisihao commented Aug 10, 2026

Copy link
Copy Markdown
Owner

No description provided.

lisihao and others added 30 commits June 5, 2026 10:56
Polish AI influence report wording and live sanitization
…electors over generic submit buttons to avoid clicking modal popups
… proof file is absent and UI mode is not required
来源: 2026-06-09 架构审计 (~/.solar/reports/2026-06-09-solar-harness-architecture-audit.md)

D1 定时任务管线复活:
- 6 个 run_*.sh 的 ${SOLAR_REPO} unbound 改为 BASH_SOURCE 自举 (launchd 环境无该变量即炸)
- run_knowledge_semantic_supervised.sh 补 SOLAR_HOME/SOLAR_KNOWLEDGE_DIR 默认值
- 新增 scripts/lib/scheduled-task.sh: 失败统一落账 state/scheduled-task-failures.jsonl
- ALLOW_PARTIAL_SUCCESS 默认 1→0, 失败不再洗成 exit 0

D2 告警接线:
- coordinator-watchdog.sh 新增 notify_event(): osascript 通知 + events/all.jsonl 落账
- coordinator/pane 两处熔断分支接通知 (此前只写日志, 死了没人知道)
- 新增 staleness_sentinel.py: 监控 telemetry 与定时任务产出 mtime + 24h 失败回看,
  超阈值通知 (24h 去重), launchd 每 30min (com.solar.staleness-sentinel)

D3 派发失败 requeue + 配额感知:
- 新增 lib/dispatch-requeue.sh: 持久重排队 (退避 60s*2^n, 上限 5 次后转人工+通知)
- dispatch_to_pane 失败路径接 requeue (此前 3 次 tmux 重试失败任务直接丢失)
- dispatch_quota_gate: 配额缓存明确耗尽时延迟入队 (fail-open, N/A 放行)
- 主循环 %10 分支消费到期重试项

验证: bash -n 全过; 净化环境实跑无 shell 错; 人为失败测试 exit 非零+落账;
requeue 退避/熔断/取件/quota gate 矩阵测试全过; ledger 注错测试通过;
test-dispatch.sh 11 PASS (2 FAIL 为预先存在的过时断言, 与本次无关)

注: run_tech_hotspot_radar.sh / run_youtube_daily_previous_day_collect.sh /
lib/evidence_ledger.py 同样已修复但混有本分支前序未提交改动, 留待分支统一收口
lisihao added 30 commits August 9, 2026 12:28
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant