diff --git a/CHANGELOG.md b/CHANGELOG.md index b18f4d5dc1..6d5baeb4db 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -8,6 +8,7 @@ ### 新增 - 专家市场默认打开「精选」,用书签标出接入市场前随安装包提供的 22 个专家模板 +- 通道卡片与 `/api/health` 展示每个通道的入站、出站、重连和丢弃计数;记忆存储页显示 checkpoint 保留窗口与 VACUUM 提示 ### 修复 - 关机时会等正在进行的数据库查询结束再关闭连接,避免 Windows 上测试进程被访问冲突打断 @@ -20,6 +21,7 @@ - 供应商的默认模型可以清空并保持为空。对话里的模型名与选择时一致,能区分不同厂家的同名模型 - 服务端升级后,登录页和已登录页面都会提示刷新前端。PostgreSQL 备份缺少客户端工具时会说明原因,Docker 镜像带上该工具 - 专家配置文件改为文本框编辑,打开时不再停在编辑器加载中 +- 飞书通道注册时改用持久化去重库,通道重建或服务重启后仍能识别重投事件 ## [1.0.3b4] - 2026-11-06 diff --git a/dashboard/src/api/modules/memoryDashboard.ts b/dashboard/src/api/modules/memoryDashboard.ts index 2caa8060d0..53647b34d9 100644 --- a/dashboard/src/api/modules/memoryDashboard.ts +++ b/dashboard/src/api/modules/memoryDashboard.ts @@ -456,6 +456,11 @@ export interface MemoryStorageInspection { wal_size: number | null; reclaimable_bytes: number | null; auto_vacuum_enabled: boolean | null; + checkpoint_retention_enabled?: boolean | null; + checkpoint_keep_last?: number | null; + checkpoint_keep_days?: number | null; + checkpoint_backup_before_prune?: boolean | null; + checkpoint_vacuum_recommended?: boolean | null; tables: { table: string; exists: boolean; diff --git a/dashboard/src/locales/en.json b/dashboard/src/locales/en.json index ea22be2fee..39d1a58c7b 100644 --- a/dashboard/src/locales/en.json +++ b/dashboard/src/locales/en.json @@ -2656,6 +2656,8 @@ "configSaved": "Configuration saved successfully", "configFailed": "Failed to save configuration", "channelEnabled": "Channel enabled", + "lastInbound": "Last inbound {{time}}", + "channelHealth": "Reconnects {{reconnect}}, dropped stale {{stale}}, duplicate {{duplicate}}", "channelDisabled": "Channel disabled", "channelType": "Channel Type", "status": "Status", @@ -4990,6 +4992,10 @@ "inspectionTimingHint": "Server-side read-only inspection time, excluding network transfer. Data size, cache and concurrent work affect it; this is not a conversation or model benchmark. Count queries have a combined 5-second budget; timed-out counts are not shown as zero.", "sharedCheckpointHint": "Multiple memory namespaces or experts share this file. Session checkpoints cannot be safely attributed, so only the current expert’s memory counts are shown. File maintenance is disabled.", "checkpointCountHint": "Checkpoints restore sessions and are not message counts. Session IDs are deduplicated to count distinct sessions. Shared content blocks store deduplicated payloads; full-text indexes are not counted again.", + "checkpointRetention": "Checkpoint retention", + "checkpointRetentionValue": "Newest {{keep}}, age limit {{days}}. One backup is written before deletion.", + "checkpointRetentionNoDays": "none", + "vacuumRecommended": "Deleting old checkpoints does not shrink the file. Reclaimable space is large; run VACUUM on this SQLite file when the agent is idle.", "comparisonHint": "Snapshots belong to this maintenance run. Deduplication preserves history and valid memories, so row counts may remain unchanged and shared blocks may increase. Background writes may affect counts. Backups consume separate space and are excluded from reclaimed bytes.", "deleteBackup": "Delete backup", "deleteBackupTitle": "Delete this backup for the current expert?", diff --git a/dashboard/src/locales/zh.json b/dashboard/src/locales/zh.json index ed091ca3b6..763e396db7 100644 --- a/dashboard/src/locales/zh.json +++ b/dashboard/src/locales/zh.json @@ -2656,6 +2656,8 @@ "configSaved": "配置保存成功", "configFailed": "配置保存失败", "channelEnabled": "频道已启用", + "lastInbound": "最近收到 {{time}}", + "channelHealth": "重连 {{reconnect}} 次,丢弃过期 {{stale}}、重复 {{duplicate}}", "channelDisabled": "频道已停用", "channelType": "通道类型", "status": "状态", @@ -4990,6 +4992,10 @@ "inspectionTimingHint": "耗时为服务端本次只读检查时间,不含网络传输;受数据量、缓存和同时运行的任务影响,不代表对话或模型性能。条数查询合计最多运行 5 秒,超时项不显示为 0。", "sharedCheckpointHint": "当前文件由多个记忆命名空间或专家共享,无法安全归属会话检查点,仅展示当前专家的记忆分类条数。文件级维护已禁用。", "checkpointCountHint": "检查点用于恢复会话,不等于消息条数;独立会话数按检查点中的会话 ID 去重。共享内容块用于去重存储,全文索引不重复计数。", + "checkpointRetention": "检查点保留", + "checkpointRetentionValue": "最近 {{keep}} 份,天数 {{days}}。删除前会先备份一份。", + "checkpointRetentionNoDays": "不限", + "vacuumRecommended": "删除旧检查点后文件不会马上变小。可回收空间已经不小,建议在空闲时对这个 SQLite 文件执行 VACUUM。", "comparisonHint": "前后快照仅对应本次维护。去重维护保留会话历史和有效记忆,因此条数可能不变,共享内容块可能增加。后台写入也可能影响统计;备份单独占用空间,未计入释放量。", "deleteBackup": "删除备份", "deleteBackupTitle": "删除当前专家的这份备份?", diff --git a/dashboard/src/pages/Agent/Channels/components/ChannelCard.tsx b/dashboard/src/pages/Agent/Channels/components/ChannelCard.tsx index ed45a5fae7..93555dc800 100644 --- a/dashboard/src/pages/Agent/Channels/components/ChannelCard.tsx +++ b/dashboard/src/pages/Agent/Channels/components/ChannelCard.tsx @@ -150,6 +150,19 @@ export function ChannelCard({

{renderStatusBadge()}
+ {runtime?.last_inbound_at ? ( +

+ {t("channels.lastInbound", { + time: new Date(runtime.last_inbound_at * 1000).toLocaleString(), + })} + {" · "} + {t("channels.channelHealth", { + reconnect: runtime.reconnect_count ?? 0, + stale: runtime.dropped_stale ?? 0, + duplicate: runtime.dropped_duplicate ?? 0, + })} +

+ ) : null} ); } diff --git a/dashboard/src/pages/Agent/Channels/useChannels.ts b/dashboard/src/pages/Agent/Channels/useChannels.ts index 6ff418dff1..9e2d1bc05a 100644 --- a/dashboard/src/pages/Agent/Channels/useChannels.ts +++ b/dashboard/src/pages/Agent/Channels/useChannels.ts @@ -26,6 +26,12 @@ export interface ChannelRow { connected: boolean; error?: string | null; updated_at?: number; + last_inbound_at?: number | null; + last_outbound_at?: number | null; + reconnect_count?: number; + last_reconnect_at?: number | null; + dropped_duplicate?: number; + dropped_stale?: number; }; } diff --git a/dashboard/src/pages/Agent/Memory/MemoryStorage.tsx b/dashboard/src/pages/Agent/Memory/MemoryStorage.tsx index d18e867815..9db9490df8 100644 --- a/dashboard/src/pages/Agent/Memory/MemoryStorage.tsx +++ b/dashboard/src/pages/Agent/Memory/MemoryStorage.tsx @@ -300,6 +300,18 @@ export default function MemoryStorage({ agentId }: { agentId: string }) { : "common.disabled", ), }, + { + key: "retention", + label: t("memory.storage.checkpointRetention"), + children: inspection.checkpoint_retention_enabled + ? t("memory.storage.checkpointRetentionValue", { + keep: inspection.checkpoint_keep_last ?? "—", + days: + inspection.checkpoint_keep_days ?? + t("memory.storage.checkpointRetentionNoDays"), + }) + : t("common.disabled"), + }, ] : []), ]} @@ -309,6 +321,13 @@ export default function MemoryStorage({ agentId }: { agentId: string }) { {t("memory.storage.estimateHint")} )} + {inspection.checkpoint_vacuum_recommended && ( + + )} {inspection.blocked_reason && ( dict[str, Any]: agents = ( len(server.app_runtime.agent_registry.list_rows()) if server.app_runtime is not None else 0 ) + channels: list[dict[str, Any]] = [] + runtime = getattr(server, "app_runtime", None) + gateway = getattr(runtime, "gateway", None) if runtime is not None else None + report = getattr(gateway, "channel_health", None) + if callable(report): + channels = report() return { "ok": True, "started_at": server._started_at, "db": bound, "users_loaded": users, "agents_running": agents, + "channels": channels, } diff --git a/src/octop/api/routers/memory_storage.py b/src/octop/api/routers/memory_storage.py index 7aa4f5f93f..02ed1be611 100644 --- a/src/octop/api/routers/memory_storage.py +++ b/src/octop/api/routers/memory_storage.py @@ -73,6 +73,11 @@ class StorageInspection(BaseModel): wal_size: int | None = None reclaimable_bytes: int | None = None auto_vacuum_enabled: bool | None = None + checkpoint_retention_enabled: bool | None = None + checkpoint_keep_last: int | None = None + checkpoint_keep_days: int | None = None + checkpoint_backup_before_prune: bool | None = None + checkpoint_vacuum_recommended: bool | None = None page_size: int | None = None total_pages: int | None = None freelist_pages: int | None = None diff --git a/src/octop/infra/agents/memory/sqlite_storage.py b/src/octop/infra/agents/memory/sqlite_storage.py index 36cfaad5bb..8993d9f727 100644 --- a/src/octop/infra/agents/memory/sqlite_storage.py +++ b/src/octop/infra/agents/memory/sqlite_storage.py @@ -2,6 +2,7 @@ from __future__ import annotations +import os import sqlite3 import time from contextlib import closing @@ -23,6 +24,49 @@ ) CHECKPOINT_TABLES = ("checkpoints", "writes", "hm_checkpoint_blobs") COUNT_BUDGET_SECONDS = 5.0 +_VACUUM_RECOMMEND_BYTES = 32 * 1024 * 1024 + + +def _env_flag(name: str, default: bool) -> bool: + raw = os.environ.get(name) + if raw is None or not str(raw).strip(): + return default + return str(raw).strip().lower() not in {"0", "false", "off", "no"} + + +def _optional_positive_int(name: str) -> int | None: + if name not in os.environ: + return None + raw = os.environ.get(name, "").strip() + if not raw or raw == "0": + return None + try: + value = int(raw) + except ValueError: + return None + return value if value > 0 else None + + +def _checkpoint_policy(reclaimable_bytes: int) -> dict[str, Any]: + """Retention window shown on the storage page. + + Matches idle maintenance in octop-harness: keep the newest 20 checkpoints + unless ``OCTOP_CHECKPOINT_RETENTION`` is turned off. + """ + keep_last = _optional_positive_int("OCTOP_CHECKPOINT_KEEP_LAST") + keep_days = _optional_positive_int("OCTOP_CHECKPOINT_KEEP_DAYS") + if not _env_flag("OCTOP_CHECKPOINT_RETENTION", True): + keep_last = None + keep_days = None + elif "OCTOP_CHECKPOINT_KEEP_LAST" not in os.environ: + keep_last = 20 + return { + "checkpoint_retention_enabled": keep_last is not None or keep_days is not None, + "checkpoint_keep_last": keep_last, + "checkpoint_keep_days": keep_days, + "checkpoint_backup_before_prune": _env_flag("OCTOP_CHECKPOINT_BACKUP_BEFORE_PRUNE", True), + "checkpoint_vacuum_recommended": reclaimable_bytes >= _VACUUM_RECOMMEND_BYTES, + } def inspect_sqlite_storage(backend: Any, *, shared: bool) -> dict[str, Any]: @@ -64,6 +108,7 @@ def inspect_sqlite_storage(backend: Any, *, shared: bool) -> dict[str, Any]: "file_check_ms": (time.perf_counter() - started) * 1000, "namespace": namespace, "checkpoints_available": not shared, + **_checkpoint_policy(free * page_size), } deadline = time.perf_counter() + COUNT_BUDGET_SECONDS conn.set_progress_handler(lambda: int(time.perf_counter() >= deadline), 1000) diff --git a/src/octop/infra/gateway/gateway.py b/src/octop/infra/gateway/gateway.py index 9e00aea21b..e47cd481d9 100644 --- a/src/octop/infra/gateway/gateway.py +++ b/src/octop/infra/gateway/gateway.py @@ -240,12 +240,35 @@ def runtime_status_to_dict( "error": error, "updated_at": status.updated_at, } + channel = self._channel_manager.get_channel(channel_id) if self._channel_manager else None + metrics = getattr(channel, "health_metrics", None) + if callable(metrics): + out.update(metrics()) if status.lease_owner: out["lease_owner"] = status.lease_owner if status.lease_until is not None: out["lease_until"] = status.lease_until return out + def channel_health(self) -> list[dict[str, Any]]: + """Per-channel liveness counters. Message bodies are not included.""" + if self._channel_manager is None: + return [] + rows: list[dict[str, Any]] = [] + for channel_id in self._channel_manager.channel_ids: + channel = self._channel_manager.get_channel(channel_id) + status = self.get_runtime_status(channel_id) + metrics = getattr(channel, "health_metrics", None) + row: dict[str, Any] = { + "id": channel_id, + "type": getattr(channel, "channel_type", ""), + "connected": bool(status.connected) if status is not None else False, + } + if callable(metrics): + row.update(metrics()) + rows.append(row) + return rows + async def boot(self, *, serve_background: bool = True) -> None: from octop.infra.gateway.process.channel_thinking import ( install_channel_thinking_clean, @@ -814,6 +837,8 @@ async def _register_channel(self, row: ChannelRow) -> None: if not self._channel_manager or not self._processor: return config = self._config_from_row(row) + if row.kind == "feishu": + config = self._with_feishu_dedup_path(config) from octop.infra.agents.teams.service import is_team_agent # noqa: PLC0415 team_host = False @@ -882,6 +907,25 @@ def _config_from_row(self, row: ChannelRow) -> dict[str, Any]: ) return raw + def _with_feishu_dedup_path(self, config: dict[str, Any]) -> dict[str, Any]: + """Point Feishu dedup at this install's data directory. + + The channel library only opens a SQLite file when a path is set. + Restart and re-register then still recognize message ids Feishu redelivers. + An explicit ``dedup_db_path`` in the channel config is left as-is. + """ + if str(config.get("dedup_db_path") or "").strip(): + return config + paths = getattr(self._agent_manager, "_paths", None) + root = getattr(paths, "root", None) + if root is None: + from octop.infra.utils.paths import PathLayout # noqa: PLC0415 + + root = PathLayout.from_env().root + updated = dict(config) + updated["dedup_db_path"] = str(root / "feishu-dedup.sqlite") + return updated + async def _unregister(self, channel_id: str) -> None: """Disconnect locally, then release only the lease this process holds.""" if self._channel_manager is not None: diff --git a/tests/unit/gateway/test_gateway_runtime_status.py b/tests/unit/gateway/test_gateway_runtime_status.py index e2190279ec..36ef623393 100644 --- a/tests/unit/gateway/test_gateway_runtime_status.py +++ b/tests/unit/gateway/test_gateway_runtime_status.py @@ -69,6 +69,9 @@ async def test_register_success_sets_runtime_connected(tmp_path: Path) -> None: assert status.reason is None assert gw._channel_manager.add_channel.await_args is not None assert gw._channel_manager.add_channel.await_args.kwargs["processor"] is not gw._processor + assert gw._channel_manager.add_channel.await_args.args[1]["dedup_db_path"] == str( + tmp_path / "feishu-dedup.sqlite" + ) @pytest.mark.asyncio