"""自动化与分诊收件箱(P1 第 2 项)。 与 host.py 现有 automations / run-due 的关系(先说清楚，避免 两套调度打架)： - host.py 的自动化存在 host state JSON 文件里(字段 name / task / repo / interval / last_run)，run-due 按 now - last_run >= interval 判断到期，到期派发一个 host 任务 跑一轮，跑完才回写 last_run。daemon tick 与 crontab 都只 调这一个 run-due，它是 host 侧唯一的执行器。 - 本模块是云端控制面的另一套登记口径：自动化定义存在控制面 数据库里(automations 表，带 project_id)，不是 host state 文件。两者存储不相交，同一个自动化绝不会同时出现在两边， 所以本批不存在双重执行。 - 本批不给云端自动化加执行器、不动 host.py 的 run-due 区： host 的 run-due 跑的是 host 任务模型(codex 协议线程)，云 端任务模型(POST /tasks 建、worker 认领)是另一条链路，把 云端定义硬塞进 host run-due 会造出第二个调度器。云端执行 器是后续一步，它必须走 claim_due_automation 先抢单(条件 UPDATE，rowcount 为 1 才算抢到)，抢到后再建云端任务、再 记 automation_runs 一行；抢不到的执行器绝不能建任务。这 个 claim 门就是防重复执行的唯一关口，比 host run-due 的 “跑完才写 last_run”更严(中途被杀也不会在当间隔内重跑， 代价是崩掉的那次要等到下个间隔，见 claim 文档)。 - 分诊收件箱是只读视图：到期自动化、失败或中断的任务、待 处理的后续消息、最近的自动化产出记录。它不跑模型、不改 任务状态，只把待处理项聚到一处。 通知接入(本批结论：遗留，不新写通知代码)： - server 没有接 webhooks.NotificationService(它只有可注入 sender 的占位实现，server docstring 也写明未接成)；真正 的终态派发是 worker 走 cloud/push.py dispatch_event(订阅 制，completed / failed / goal-unmet 等终态才发)。 - 自动化本身没有终态事件可派：定义创建不是完成，取消不是 失败；硬在创建/取消时发通知是伪造成功。本批只记审计 (automation_created / automation_cancelled)。等执行器 落地后，它建的云端任务自然走 worker 现有终态派发(任务 id 记在 automation_runs.task_id 里，派发与任务绑定，不用 为自动化另起一套通知系统)。 本模块只依赖标准库，不 import cloud 其它模块，避免循环 import；表迁移写法仿 peer_messages(CREATE TABLE IF NOT EXISTS + PRAGMA 查列补齐，SQLite 与 Postgres 适配层同款)。 """ import time import uuid MIN_INTERVAL_SECONDS = 60 MAX_INTERVAL_SECONDS = 30 * 86400 MAX_NAME_LEN = 100 MAX_PROMPT_LEN = 20000 MAX_REPO_URL_LEN = 1000 TRIAGE_LIMIT = 50 ACTIVE = "active" CANCELLED = "cancelled" def row_get(row, name): """行取值容错：sqlite3.Row 缺列报 IndexError、dict 缺键报 KeyError、空行报 TypeError，一律回 None(老库行缺列时 读自动化列表不能因此炸掉整页)。""" try: return row[name] except (KeyError, IndexError, TypeError): return None AUTOMATION_COLUMNS = ( "id", "project_id", "name", "prompt", "repo_url", "interval_seconds", "status", "last_run_at", "last_claimed_at", "created_by", "created_at", "cancelled_at") RUN_COLUMNS = ( "id", "automation_id", "project_id", "task_id", "status", "detail", "created_at") def _row_dict(row, columns): """把行统一成可按名取值的形态：dict 与 sqlite3.Row 原样 用；裸 sqlite3 连接给的是 tuple(按 SELECT 列顺序)， 按建表列顺序 zip 成 dict——直接调模块的调用方不一定设 row_factory，模块不能因此在取 row[id] 处炸掉。""" if row is None or isinstance(row, dict): return row try: row["id"] return row except TypeError: return dict(zip(columns, row)) def _existing_columns(conn, table): try: return {row[1] for row in conn.execute( "PRAGMA table_info(%s)" % table).fetchall()} except Exception: return set() def ensure_automations_tables(conn): """幂等建自动化表与产出记录表，并补齐新增列。 automations：一条定义；status 只取 active / cancelled， 取消是软删除(行保留、审计可查，名字可重用)。 automation_runs：一次执行产出；task_id 为空表示执行器 抢单后尚未建任务(只应短暂存在，执行器建任务后补记)。 调用方在写成功后 commit(server 的 handler 与 peer 模式 同款：建表后由业务写操作一并 commit)。""" conn.execute( "CREATE TABLE IF NOT EXISTS automations(" "id TEXT PRIMARY KEY, " "project_id TEXT, " "name TEXT, " "prompt TEXT, " "repo_url TEXT, " "interval_seconds INTEGER, " "status TEXT, " "last_run_at DOUBLE PRECISION, " "last_claimed_at DOUBLE PRECISION, " "created_by TEXT, " "created_at DOUBLE PRECISION, " "cancelled_at DOUBLE PRECISION)" ) conn.execute( "CREATE TABLE IF NOT EXISTS automation_runs(" "id TEXT PRIMARY KEY, " "automation_id TEXT, " "project_id TEXT, " "task_id TEXT, " "status TEXT, " "detail TEXT, " "created_at DOUBLE PRECISION)" ) existing = _existing_columns(conn, "automations") for col, ddl in ( ("repo_url", "TEXT"), ("last_claimed_at", "DOUBLE PRECISION"), ("cancelled_at", "DOUBLE PRECISION")): if col not in existing: try: conn.execute( "ALTER TABLE automations ADD COLUMN %s %s" % (col, ddl)) except Exception: pass existing_runs = _existing_columns(conn, "automation_runs") for col, ddl in ( ("task_id", "TEXT"), ("detail", "TEXT")): if col not in existing_runs: try: conn.execute( "ALTER TABLE automation_runs ADD COLUMN %s %s" % (col, ddl)) except Exception: pass conn.execute( "CREATE INDEX IF NOT EXISTS idx_automations_project" " ON automations(project_id, status)") conn.execute( "CREATE INDEX IF NOT EXISTS idx_automation_runs" " ON automation_runs(automation_id, created_at)") conn.execute( "CREATE INDEX IF NOT EXISTS idx_automation_runs_project" " ON automation_runs(project_id, created_at)") def next_run_at(auto, now=None): """下次到期时刻：从未跑过时就是创建时间(建完即到期， 执行器下一轮即可认领)；跑过时为上次执行 + 间隔。""" last = row_get(auto, "last_run_at") interval = row_get(auto, "interval_seconds") or 0 if last is None: return row_get(auto, "created_at") return last + interval def is_due(auto, now=None): """是否到期：仅 active 的定义可到期；now 缺省取当前时间。 到期边界用 >=，与 host run-due 的 now - last_run >= interval 同款(含边界那一秒)。""" if row_get(auto, "status") != ACTIVE: return False if now is None: now = time.time() nxt = next_run_at(auto) if nxt is None: return True return now >= nxt def automation_to_dict(row, now=None): row = _row_dict(row, AUTOMATION_COLUMNS) nxt = next_run_at(row) return { "id": row["id"], "project_id": row["project_id"], "name": row["name"], "prompt": row["prompt"], "repo_url": row_get(row, "repo_url"), "interval_seconds": row["interval_seconds"], "status": row["status"], "last_run_at": row_get(row, "last_run_at"), "last_claimed_at": row_get(row, "last_claimed_at"), "created_by": row["created_by"], "created_at": row["created_at"], "cancelled_at": row_get(row, "cancelled_at"), "next_run_at": nxt, "due": is_due(row, now), } def _validate_fields(name, prompt, interval_seconds, repo_url): if not name or not isinstance(name, str) or not name.strip(): raise ValueError("name required") name = name.strip() if len(name) > MAX_NAME_LEN: raise ValueError("name too long") if not prompt or not isinstance(prompt, str) or not prompt.strip(): raise ValueError("prompt required") prompt = prompt.strip() if len(prompt) > MAX_PROMPT_LEN: raise ValueError("prompt too long") if isinstance(interval_seconds, bool) \ or not isinstance(interval_seconds, int): raise ValueError("interval_seconds must be an integer") if interval_seconds < MIN_INTERVAL_SECONDS: raise ValueError( "interval_seconds must be at least %d" % MIN_INTERVAL_SECONDS) if interval_seconds > MAX_INTERVAL_SECONDS: raise ValueError("interval_seconds too large") if repo_url is not None: if not isinstance(repo_url, str): raise ValueError("repo_url must be a string") repo_url = repo_url.strip() or None if repo_url is not None and len(repo_url) > MAX_REPO_URL_LEN: raise ValueError("repo_url too long") return name, prompt, interval_seconds, repo_url def create_automation(conn, project_id, name, prompt, created_by, interval_seconds=86400, repo_url=None, now=None): """创建自动化定义：同项目内 active 的名字唯一(取消过的 名字可重用，与环境名唯一规则同款语义)。成功时 commit 并返回 dict；校验失败抛 ValueError(调用方转 400)。""" name, prompt, interval_seconds, repo_url = _validate_fields( name, prompt, interval_seconds, repo_url) if now is None: now = time.time() ensure_automations_tables(conn) existing = conn.execute( "SELECT id FROM automations WHERE project_id = ?" " AND name = ? AND status = ?", (project_id, name, ACTIVE)).fetchone() if existing is not None: raise ValueError("automation name already exists in this project") auto_id = uuid.uuid4().hex conn.execute( "INSERT INTO automations(id, project_id, name, prompt," " repo_url, interval_seconds, status, last_run_at," " last_claimed_at, created_by, created_at, cancelled_at)" " VALUES(?, ?, ?, ?, ?, ?, ?, NULL, NULL, ?, ?, NULL)", (auto_id, project_id, name, prompt, repo_url, interval_seconds, ACTIVE, created_by, now)) conn.commit() row = conn.execute( "SELECT * FROM automations WHERE id = ?", (auto_id,)).fetchone() return automation_to_dict(row, now) def list_automations(conn, project_id, now=None): """列出某项目全部自动化定义(含已取消，取消的 due 为 False)，按创建时间排序。调用方须先 ensure 表。""" if now is None: now = time.time() ensure_automations_tables(conn) rows = conn.execute( "SELECT * FROM automations WHERE project_id = ?" " ORDER BY created_at, id", (project_id,)).fetchall() return [automation_to_dict(r, now) for r in rows] def get_automation(conn, automation_id): """取单条定义；不存在回 None(调用方转 404)。""" ensure_automations_tables(conn) row = conn.execute( "SELECT * FROM automations WHERE id = ?", (automation_id,) ).fetchone() if row is None: return None return automation_to_dict(row) def cancel_automation(conn, automation_id, now=None): """取消自动化(软删除)：仅 active 的行能取消。 返回 (ok, auto)：ok 为 True 时 auto 是取消后的 dict； 行不存在时 (False, None)；已取消时 (False, 旧 dict)， 调用方据此区分 404 与 409。成功时 commit。""" if now is None: now = time.time() ensure_automations_tables(conn) row = conn.execute( "SELECT * FROM automations WHERE id = ?", (automation_id,) ).fetchone() if row is None: return False, None row = _row_dict(row, AUTOMATION_COLUMNS) if row["status"] != ACTIVE: return False, automation_to_dict(row, now) cur = conn.execute( "UPDATE automations SET status = ?, cancelled_at = ?" " WHERE id = ? AND status = ?", (CANCELLED, now, automation_id, ACTIVE)) conn.commit() if cur.rowcount != 1: return False, automation_to_dict(row, now) row = conn.execute( "SELECT * FROM automations WHERE id = ?", (automation_id,) ).fetchone() return True, automation_to_dict(row, now) def due_automations(conn, project_id, now=None): """列出某项目当前到期的 active 定义(纯读，不抢单)。 SQL 层算到期：从未跑过(last_run_at 为空)或上次执行距 今已满间隔；与 is_due 的 Python 算法同款。""" if now is None: now = time.time() ensure_automations_tables(conn) rows = conn.execute( "SELECT * FROM automations WHERE project_id = ?" " AND status = ?" " AND (last_run_at IS NULL" " OR last_run_at + interval_seconds <= ?)" " ORDER BY created_at, id", (project_id, ACTIVE, now)).fetchall() return [automation_to_dict(r, now) for r in rows] def claim_due_automation(conn, automation_id, now=None): """抢单：未来执行器的唯一入口。 条件 UPDATE 只有一行能成功：同一自动化在同一间隔内， 两个并发执行器只有一个拿到 rowcount 1，另一个拿到 0 就必须放弃建任务——这是防重复执行的硬门。抢单即把 last_run_at 写成 now(不是等执行完才写，与 host run-due 相反)：好处是崩掉的执行不会在当间隔内被另一执行器重 跑；代价是抢单后崩掉的那次产出会缺失，直到下个间隔才 补跑。执行器抢单成功后必须调 record_automation_run 记 产出，否则分诊收件箱里这次执行不可见。成功时 commit。 返回 True 表示抢到，False 表示不存在、已取消或未到期。 """ if now is None: now = time.time() ensure_automations_tables(conn) cur = conn.execute( "UPDATE automations SET last_run_at = ?," " last_claimed_at = ?" " WHERE id = ? AND status = ?" " AND (last_run_at IS NULL" " OR last_run_at + interval_seconds <= ?)", (now, now, automation_id, ACTIVE, now)) conn.commit() return cur.rowcount == 1 def record_automation_run(conn, automation_id, project_id, task_id=None, status="created", detail="", now=None): """记一次自动化产出：执行器建任务后调它，task_id 为建好 的云端任务 id；status 取 created / completed / failed 等 执行器自定的终态词(本模块不校验任务状态机，只原样存)。 自动化不存在时抛 ValueError(执行器绝不能给不存在的 定义记产出)。成功时 commit 并返回 run dict。""" if now is None: now = time.time() ensure_automations_tables(conn) auto = conn.execute( "SELECT id FROM automations WHERE id = ?", (automation_id,)).fetchone() if auto is None: raise ValueError("automation not found") run_id = uuid.uuid4().hex conn.execute( "INSERT INTO automation_runs(id, automation_id, project_id," " task_id, status, detail, created_at)" " VALUES(?, ?, ?, ?, ?, ?, ?)", (run_id, automation_id, project_id, task_id, status, detail or "", now)) conn.commit() return { "id": run_id, "automation_id": automation_id, "project_id": project_id, "task_id": task_id, "status": status, "detail": detail or "", "created_at": now, } def list_automation_runs(conn, automation_id, limit=TRIAGE_LIMIT): """列某自动化的最近产出记录，按时间倒序。""" ensure_automations_tables(conn) rows = conn.execute( "SELECT * FROM automation_runs WHERE automation_id = ?" " ORDER BY created_at DESC, id LIMIT ?", (automation_id, limit)).fetchall() out = [] for raw in rows: r = _row_dict(raw, RUN_COLUMNS) out.append({ "id": r["id"], "automation_id": r["automation_id"], "project_id": r["project_id"], "task_id": row_get(r, "task_id"), "status": r["status"], "detail": row_get(r, "detail"), "created_at": r["created_at"], }) return out def triage_inbox(conn, project_id, now=None): """分诊收件箱(只读，不写业务行、不改任务状态)： - automations_due：当前到期的自动化定义。 - failed_tasks：失败或中断的任务(最需要人工看的项)， 按更新时间倒序。 - pending_followups：仍 pending 的后续消息(含任务间 消息排队的那部分，带 sender_task_id)，按创建时间排序； followups 表不存在时该段为空(不因此炸掉整页)。 - recent_runs：最近的自动化产出记录，带对应任务的当前 状态(任务删了则 task_status 为 None，记录本身保留)。 调用方(server handler)须先 ensure 自动化表与 followups 表；本函数内 followups 缺表时只吞缺表错，其它错照常抛 出(调用方转 502，不伪造一页空收件箱)。""" if now is None: now = time.time() ensure_automations_tables(conn) due = due_automations(conn, project_id, now) task_rows = conn.execute( "SELECT id, status, prompt, created_at, updated_at," " error_text FROM tasks WHERE project_id = ?" " AND status IN ('failed', 'interrupted')" " ORDER BY updated_at DESC, id LIMIT ?", (project_id, TRIAGE_LIMIT)).fetchall() failed = [] for raw in task_rows: r = _row_dict(raw, ("id", "status", "prompt", "created_at", "updated_at", "error_text")) failed.append({ "id": r["id"], "status": r["status"], "prompt": (r["prompt"] or "")[:300], "created_at": r["created_at"], "updated_at": r["updated_at"], "error_text": (row_get(r, "error_text") or "")[:300], }) try: fu_rows = conn.execute( "SELECT f.id, f.task_id, f.body, f.mode, f.status," " f.sender_task_id, f.created_at FROM task_followups f" " JOIN tasks t ON t.id = f.task_id" " WHERE f.project_id = ? AND f.status = 'pending'" " ORDER BY f.created_at, f.id LIMIT ?", (project_id, TRIAGE_LIMIT)).fetchall() pending = [] for raw in fu_rows: r = _row_dict(raw, ("id", "task_id", "body", "mode", "status", "sender_task_id", "created_at")) pending.append({ "id": r["id"], "task_id": r["task_id"], "body": (r["body"] or "")[:300], "mode": r["mode"], "status": r["status"], "sender_task_id": row_get(r, "sender_task_id"), "created_at": r["created_at"], }) except Exception as exc: msg = str(exc) if "no such table" in msg or "does not exist" in msg: pending = [] else: raise run_rows = conn.execute( "SELECT r.id, r.automation_id, r.task_id, r.status," " r.detail, r.created_at, t.status AS task_status" " FROM automation_runs r" " LEFT JOIN tasks t ON t.id = r.task_id" " WHERE r.project_id = ?" " ORDER BY r.created_at DESC, r.id LIMIT ?", (project_id, TRIAGE_LIMIT)).fetchall() runs = [] for raw in run_rows: r = _row_dict(raw, ("id", "automation_id", "task_id", "status", "detail", "created_at", "task_status")) runs.append({ "id": r["id"], "automation_id": r["automation_id"], "task_id": row_get(r, "task_id"), "status": r["status"], "detail": row_get(r, "detail"), "created_at": r["created_at"], "task_status": row_get(r, "task_status"), }) return { "project_id": project_id, "automations_due": due, "failed_tasks": failed, "pending_followups": pending, "recent_runs": runs, "counts": { "automations_due": len(due), "failed_tasks": len(failed), "pending_followups": len(pending), "recent_runs": len(runs), }, }