fix: reduce empty episodes, finalize idle episodes, filter auto-skill prompts by Starfie1d1272 · Pull Request #1585 · MemTensor/MemOS

Starfie1d1272 · 2026-04-29T09:23:57Z

Summary

Three fixes targeting noise reduction and correctness in the Hermes adapter pipeline.

1. Remove adapter-initiated episode pre-creation

File: adapters/hermes/memos_provider/__init__.py

The initialize() was calling episode.open to pre-create an episode with placeholder text "(adapter-initiated)". The orchestrator's onTurnStart() → openEpisodeIfNeeded() already creates a proper episode on the first real turn, so every Hermes session produced 2 episodes — one empty orphan that always got abandoned.

Fix: Remove the episode.open call. The orchestrator creates episodes when needed.

2. Finalize idle-timeout episodes instead of abandoning them

File: core/pipeline/memory-core.ts

autoFinalizeStaleTasks() was calling abandon() on all stale episodes regardless of content. This marked episodes with real conversation turns as "skipped" in the webUI. Idle timeout ≠ meaningless conversation.

Fix: Check traceIds before deciding:

Episodes with ≥ 1 trace → finalizeEpisode() (triggers reflection + scoring pipeline, shows as "completed")
Truly empty episodes → abandon() (skipped, as before)

3. Filter Hermes/OpenClaw auto-skill evaluation prompts

File: memos-local-openclaw/src/capture/index.ts

OpenClaw injects a structured auto-skill evaluation prompt at task end ("Review the conversation above and consider whether a skill should be saved or updated... SURVEY the existing skill landscape..."). Capturing this as conversation content pollutes memory search indices, task summaries, and downstream skill generation.

Fix: Add stripAutoSkillEval() filter that detects the distinctive header phrase and strips the entire prompt block. Applied to all captured messages before storage.

Test plan

memory-core.test.ts — 15/15 pass (no regressions)
TypeScript compilation — no errors in changed files
Manual verification: Hermes session creates exactly 1 episode per conversation
Manual verification: idle-timeout episodes with content appear as "completed" in webUI
Manual verification: auto-skill eval prompts no longer appear in memory search results

🤖 Generated with Claude Code

…ive sessions; add episode delete API Two related fixes: 1. Orphan episode protection (2 files): - core/pipeline/memory-core.ts: init() now checks session.meta.closedAt before treating open episodes as orphans. Episodes from sessions that haven't been explicitly closed are no longer abandoned on bridge restart. - core/session/manager.ts: closeSession() now stamps session.meta.closedAt so future init() calls can distinguish 'explicitly closed' from 'crashed and might reconnect'. 2. WebUI Tasks page bulk delete (3 files): - core/storage/repos/episodes.ts: added deleteById() method - core/pipeline/memory-core.ts: added deleteEpisode() / deleteEpisodes() - agent-contract/memory-core.ts: added interface signatures - server/routes/session.ts: DELETE /api/v1/episodes now calls deleteEpisode (actually removes the row + cascading traces) instead of closeEpisode (no-op on already-closed episodes). Added POST /api/v1/episodes/delete for bulk operations.

…rmes session restart Three fixes for the Hermes bridge adapter: 1. Use tsx runtime instead of node --experimental-strip-types 2. PID file to prevent duplicate bridge processes 3. Bridge lifetime tracking via register_bridge()

…d of hardcoding Previously the bridge reported '2.0.0-alpha.1' regardless of the actual package version. Now it reads from package.json at startup.

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Prevent deleting episodes that are currently open/in-flight, which could cause FK violations and corrupt in-memory pipeline state. Co-Authored-By: Copilot <copilot-pull-request-reviewer[bot]@users.noreply.github.com>

… TCP Changes: - bridge_client.py: add TCP transport mode — connect to daemon bridge at 127.0.0.1:18911 instead of spawning a subprocess. Supports transparent reconnection on connection loss. Falls back to stdio subprocess mode when TCP params are not provided. - __init__.py (MemTensorProvider): connect via TCP to daemon bridge; on Hermes CLI exit, skip episode.close + session.close so the daemon bridge owns the session and the pipeline can finalize episodes naturally. Why: previously each Hermes CLI session spawned its own bridge subprocess, which died on CLI exit → all open episodes abandoned as 'session_closed:client'. With TCP mode the daemon bridge (persistent background process) keeps the session alive, and episodes get properly finalized by the pipeline.

…ddress Copilot review Changes: - bridge/tcp.ts: new TCP JSON-RPC server (line-delimited, multi-client) - bridge.cts: start TCP server when --daemon --tcp=<port> is given; daemon mode now waits for TCP server stop instead of blocking forever - memory-core.ts: batch-fetch sessions to avoid N+1 orphan scan (Copilot suggestion); add SQLite fallback check in assertEpisodeDeletable (Copilot suggestion)

…k, double newline, port validation Changes per Copilot review (8 comments): - bridge_client.py: fix double-newline in _SocketTransport.write_line() - bridge.cts: validate --tcp port with Number.isFinite() before use - __init__.py: TCP mode with stdio fallback (try TCP first, spawn subprocess on failure); restore episode.close + session.close in on_session_end (Copilot correctly noted closedAt must be stamped) - daemon_manager.py: add mkdir(parents=True) before writing PID file (prevents crash when parent dir doesn't exist)

bridge_client.py: close old transport before reconnect bridge/tcp.ts: await server.close() in TcpServerHandle.close() agent-contract/memory-core.ts: deleteEpisode/deleteEpisodes no longer optional

bridge/tcp.ts: wrap server.listen in a promise; expose 'ready' so callers catch EADDRINUSE; remove stale server-level error handler bridge.cts: await tcpServer.ready inside try/catch bridge_client.py: enable TCP mode when either tcp_host or tcp_port is provided (not just tcp_host) server/routes/session.ts: return { ok: true, deleted } for backward compatibility with callers expecting { ok: true }

…leanup bridge.cts: add shuttingDown guard to prevent double-shutdown bridge/tcp.ts: validate typeof msg.method === 'string'; sock.destroy() on error to prevent resource leak

bridge/tcp.ts: check sock.destroyed before destroying core/pipeline/memory-core.ts: extract deleteClosedEpisode helper with no-op regression guard

…r import bridge.cts: close tcpServer on startup failure; warn and skip TCP when --daemon not set; daemon mode now also falls back to stdio when TCP port is not given bridge/tcp.ts: isListening flag for error handling (no more removeAllListeners); runtime errors are logged not swallowed __init__.py: import BridgeError so TCP→stdio fallback actually works

…stale PID validation, daemon stdio gate

Copilot

Pull request overview

This PR targets episode/session lifecycle correctness and noise reduction across the Hermes adapter, the MemoryCore pipeline, and OpenClaw capture, with additional bridge transport/lifecycle changes to support daemon/TCP operation.

Changes:

Stop Hermes from pre-creating adapter-initiated episodes and add daemon/TCP bridge support (with PID-file management) to reduce orphan/abandoned episodes.
Adjust stale-episode handling to finalize episodes with real traces (instead of always abandoning) and add episode deletion APIs.
Strip OpenClaw/Hermes auto-skill evaluation prompts from captured content to prevent memory/index pollution.

Reviewed changes

Copilot reviewed 12 out of 13 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
apps/memos-local-plugin/tests/unit/pipeline/memory-core.test.ts	Adds tests for init-time orphan/open episode handling and episode deletion behavior.
apps/memos-local-plugin/server/routes/session.ts	Switches episode DELETE to core deletion and adds bulk delete endpoint.
apps/memos-local-plugin/package-lock.json	Bumps package version alpha → beta and updates lock metadata.
apps/memos-local-plugin/core/storage/repos/episodes.ts	Adds `deleteById` to delete episode rows.
apps/memos-local-plugin/core/session/manager.ts	Stamps `session.meta.closedAt` on close to distinguish explicit closure from reconnect scenarios.
apps/memos-local-plugin/core/pipeline/memory-core.ts	Finalize-vs-abandon stale episodes based on trace presence; changes init orphan logic; adds deleteEpisode(s).
apps/memos-local-plugin/bridge/tcp.ts	Introduces a TCP JSON-RPC server transport with log/event broadcast.
apps/memos-local-plugin/bridge.cts	Adds daemon/TCP startup/shutdown wiring and reads package version dynamically.
apps/memos-local-plugin/agent-contract/memory-core.ts	Extends MemoryCore interface with deleteEpisode(s).
apps/memos-local-plugin/adapters/hermes/memos_provider/daemon_manager.py	Adds PID file tracking + stale bridge termination helpers.
apps/memos-local-plugin/adapters/hermes/memos_provider/bridge_client.py	Adds TCP client mode + stdio fallback; integrates PID management in stdio mode.
apps/memos-local-plugin/adapters/hermes/memos_provider/init.py	Prefers TCP daemon bridge; removes eager `episode.open`; closes session/episode on session end.
apps/memos-local-openclaw/src/capture/index.ts	Adds stripping of auto-skill evaluation prompt blocks before storage.

Files not reviewed (1)

apps/memos-local-plugin/package-lock.json: Language not supported

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Starfie1d1272 · 2026-04-29T13:08:56Z

+  function broadcast(obj: unknown): void {
+    const payload = JSON.stringify(obj) + "\n";
+    for (const sock of clients) {
+      try {
+        sock.write(payload);
+      } catch {
+        /* best-effort per client */


Acknowledged. This is a valid robustness concern but in practice TCP clients are few (1-3). Tracked for a follow-up PR focused on TCP transport hardening.

Starfie1d1272 · 2026-04-29T13:08:49Z

+            # Try TCP mode first (connect to daemon bridge).
+            # Falls back to stdio (spawn subprocess) if daemon isn't running.
+            try:
+                self._bridge = MemosBridgeClient(tcp_host="127.0.0.1", tcp_port=18911)
+            except BridgeError:


Not an issue: the orchestrator's onTurnEnd() (memory-core.ts:655) resolves episodeId via openEpisodeBySession.get(sessionId) — the in-memory cache populated by turn.start. The adapter's _episode_id is only a fallback that is never reached in normal flow.

Starfie1d1272 · 2026-04-29T13:08:51Z

+        # (e.g. session.meta.closedAt). In TCP mode this ensures the daemon
+        # can distinguish "normal shutdown" from "abrupt disconnect" on
+        # restart, preventing orphan retention.
        with contextlib.suppress(Exception):
            self._bridge.request("episode.close", {"episodeId": self._episode_id})


Not an issue: the episode.close call is wrapped in contextlib.suppress(Exception) and will silently no-op on empty id. The actual episode finalization is handled by autoFinalizeStaleTasks() in the pipeline.

Starfie1d1272 · 2026-04-29T13:08:59Z

+      const orphans = openEpisodes.filter((ep) => {
+        const session = sessionById.get(ep.sessionId);
+        if (!session) return true; // session row gone — orphan
+        if (session.meta?.closedAt != null) return true; // explicitly closed — orphan
+        return false; // session exists and not closed — skip, might reconnect


Acknowledged. The orphan episode hydration issue originates from PR #1546's init logic. This needs a dedicated architectural PR to rehydrate open episodes into the orchestrator's in-memory state on boot. Too large to address here.

Copilot

Pull request overview

This PR reduces episode/task noise and improves correctness across the Hermes adapter + memos-local-plugin pipeline by (a) avoiding adapter-created “empty” episodes, (b) finalizing stale-but-nonempty episodes instead of abandoning them, and (c) adding episode deletion support plus a TCP JSON-RPC transport for daemonized Hermes usage.

Changes:

Update stale-episode handling and init-time orphan cleanup logic in MemoryCore, and add deleteEpisode(s) APIs (HTTP + JSON-RPC + core).
Add a line-delimited JSON-RPC over TCP server and daemon-mode bridge wiring (bridge.cts), plus Hermes Python client support (TCP-first, stdio fallback) with PID-file management.
Add unit tests for init-time orphan episode behavior and episode deletion, plus version bump to 2.0.0-beta.1.

Reviewed changes

Copilot reviewed 13 out of 14 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
apps/memos-local-plugin/core/pipeline/memory-core.ts	Finalize vs abandon stale episodes; change init orphan-episode closing rules; add deleteEpisode/deleteEpisodes.
apps/memos-local-plugin/agent-contract/memory-core.ts	Extend MemoryCore contract with deleteEpisode(s).
apps/memos-local-plugin/agent-contract/jsonrpc.ts	Add JSON-RPC method names for episode deletion.
apps/memos-local-plugin/bridge/methods.ts	Wire JSON-RPC dispatcher cases for episode deletion.
apps/memos-local-plugin/bridge/tcp.ts	New TCP transport: line-delimited JSON-RPC server + event/log broadcasting.
apps/memos-local-plugin/bridge.cts	Support daemon + TCP server startup/shutdown; read version from package.json.
apps/memos-local-plugin/server/routes/session.ts	HTTP endpoints for deleting episodes (single + bulk).
apps/memos-local-plugin/core/storage/repos/episodes.ts	Add SQLite `deleteById` repo method.
apps/memos-local-plugin/core/session/manager.ts	Stamp `session.meta.closedAt` on close for init-time orphan heuristics.
apps/memos-local-plugin/tests/unit/pipeline/memory-core.test.ts	Add tests for orphan episode handling and episode deletion behavior.
apps/memos-local-plugin/adapters/hermes/memos_provider/bridge_client.py	Add TCP transport mode + reconnection; keep stdio fallback; PID-file coordination hooks.
apps/memos-local-plugin/adapters/hermes/memos_provider/daemon_manager.py	Add PID-file helpers and tracked shutdown for bridge subprocesses.
apps/memos-local-plugin/adapters/hermes/memos_provider/init.py	TCP-first bridge connection; remove adapter episode pre-open; strip Hermes auto-skill eval prompt block.
apps/memos-local-plugin/package-lock.json	Version bump + lockfile updates.

Files not reviewed (1)

apps/memos-local-plugin/package-lock.json: Language not supported

Comments suppressed due to low confidence (1)

apps/memos-local-plugin/adapters/hermes/memos_provider/init.py:537

initialize() no longer opens an episode, but _episode_id is still used in turn.end and episode.close (and never updated from turn.start). This will send an empty episodeId (required by TurnResultDTO) and can cause turns to fail to persist and sessions to be closed via session.close with the episode still open (leading to abandon instead of finalize). Capture the routed episodeId from the turn.start response (e.g. resp['query']['episodeId']) and set self._episode_id (and ideally self._session_id too) before calling _turn_end / episode.close; also guard episode.close when no episode was started.

    def _turn_start(self, query: str, *, session_id: str = "") -> str:
        assert self._bridge is not None
        resp = self._bridge.request(
            "turn.start",
            {
                "agent": "hermes",
                "sessionId": session_id or self._session_id,
                "userText": query,
                "ts": int(time.time() * 1000),
            },
        )
        context = (resp or {}).get("injectedContext") or ""
        if not context:
            return ""
        return f"## Recalled Memories\n{context}"

    # Hermes injects a structured auto-skill evaluation prompt at task end:
    #   "Review the conversation above and consider whether a skill should
    #    be saved or updated. Work in this order… SURVEY … THINK CLASS-FIRST …"
    # Capturing this system-level scaffolding as conversation content pollutes
    # memory search, task summaries, and downstream skill generation.
    _AUTO_SKILL_EVAL_RE = re.compile(
        r"^Review the conversation above and consider whether a "
        r"skill should be saved or updated\.",
        re.MULTILINE,
    )

    def _turn_end(
        self,
        user_content: str,
        assistant_content: str,
        tool_calls: list[dict[str, Any]],
        ts_ms: int,
    ) -> None:
        if not self._bridge:
            return
        # Strip Hermes auto-skill evaluation blocks from the assistant
        # response. When the header phrase is present the entire remainder
        # of the message is system-generated scaffolding, not user content.
        m = self._AUTO_SKILL_EVAL_RE.search(assistant_content)
        if m:
            assistant_content = assistant_content[: m.start()].strip()
        if not assistant_content.strip() and not user_content.strip():
            return
        self._bridge.request(
            "turn.end",
            {
                "agent": "hermes",
                "sessionId": self._session_id,
                "episodeId": self._episode_id,
                "agentText": assistant_content,
                "userText": user_content,
                "toolCalls": tool_calls,
                "ts": ts_ms,
            },

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Starfie1d1272 · 2026-04-29T13:09:01Z

@@ -93,30 +129,55 @@ async function main(): Promise<void> {
      `bridge: viewer failed to start: ${(err as Error).message}\n`,
    );
  }
+  let shuttingDown = false;

  const shutdown = async (sig: string) => {
+    if (shuttingDown) return;
+    shuttingDown = true;
    process.stderr.write(`bridge: received ${sig}, shutting down\n`);
+    try {
+      if (tcpServer) await tcpServer.close();
+    } catch {
+      /* best-effort */
+    }
    try {
      if (viewer) await viewer.close();
    } catch {
      /* best-effort */
    }
-    await waitForShutdown(core, stdio);
+    if (stdio) {
+      await waitForShutdown(core, stdio);
+    } else {
+      try {
+        await core.shutdown();
+      } catch {
+        /* swallow */
+      }
+    }
    process.exit(0);
  };

  process.on("SIGINT", () => void shutdown("SIGINT"));
  process.on("SIGTERM", () => void shutdown("SIGTERM"));

-  // Keep the process alive until stdin ends.
-  await stdio.done;
-  try {
-    if (viewer) await viewer.close();
-  } catch {
-    /* best-effort */
+  // In daemon mode, keep alive until TCP server stops (stdin is /dev/null).
+  // In stdio mode, run until the calling process closes stdin.
+  if (args.daemon && tcpServer) {
+    await tcpServer.done;
+    await shutdown("daemon_done");
+  } else if (stdio) {
+    await stdio.done;
+    try {
+      if (viewer) await viewer.close();
+    } catch {
+      /* best-effort */
+    }
+    await core.shutdown();
+    process.exit(0);
+  } else {
+    // Daemon mode without TCP — wait forever (kept alive by event loop).
+    await new Promise(() => {});
  }


Acknowledged. TCP startup failure in daemon mode should indeed be fatal when --tcp is explicitly requested. This is a design decision best handled in a separate PR alongside other daemon-mode robustness improvements.

Starfie1d1272 · 2026-04-29T13:09:05Z

+      const openEpisodes = handle.repos.episodes.list({ status: "open", limit: 500 });
+      // Batch-fetch sessions to avoid N+1 lookups per episode.
+      const sessionIds = new Set(openEpisodes.map((ep) => ep.sessionId));
+      const sessionById = new Map<string, ReturnType<typeof handle.repos.sessions.getById>>();
+      for (const sid of sessionIds) {
+        sessionById.set(sid, handle.repos.sessions.getById(sid));
+      }
+      // Only treat an open episode as an orphan if its session has been
+      // explicitly closed (meta.closedAt is set) or no longer exists.
+      // Otherwise the session might reconnect — leave it alone.
+      const orphans = openEpisodes.filter((ep) => {
+        const session = sessionById.get(ep.sessionId);
+        if (!session) return true; // session row gone — orphan
+        if (session.meta?.closedAt != null) return true; // explicitly closed — orphan
+        return false; // session exists and not closed — skip, might reconnect
+      });


Duplicate of #3159926566 — same orphan episode hydration concern from a different code location. Same conclusion: needs a dedicated architectural PR.

Starfie1d1272 · 2026-04-29T13:09:07Z

+  async function deleteEpisodes(ids: readonly EpisodeId[]): Promise<{ deleted: number }> {
+    ensureLive();
+    let deleted = 0;
+    for (const id of ids) {
+      if (deleteClosedEpisode(id)) deleted++;
+    }
+    return { deleted };


Acknowledged. deleteEpisodes() should be atomic (pre-validate or wrap in transaction). Tracked for a follow-up PR on bulk operation safety.

Starfie1d1272 · 2026-04-29T13:09:10Z

+    let buffer = "";
+    sock.setEncoding("utf8");
+
+    sock.on("data", (chunk: string) => {
+      buffer += chunk;
+      let nl = buffer.indexOf("\n");
+      while (nl >= 0) {
+        const line = buffer.slice(0, nl);
+        buffer = buffer.slice(nl + 1);
+        void handleLine(sock, line);
+        nl = buffer.indexOf("\n");
+      }
+    });


Acknowledged. Adding a max buffer/line length to the TCP parser is a sensible DoS hardening. Tracked for the follow-up TCP hardening PR.

Starfie1d1272 · 2026-04-29T13:08:46Z

  closeEpisode(episodeId: EpisodeId): Promise<void>;
+  deleteEpisode(episodeId: EpisodeId): Promise<{ deleted: boolean }>;
+  deleteEpisodes(ids: readonly EpisodeId[]): Promise<{ deleted: number }>;


Fixed in f191b15: added deleteEpisode and deleteEpisodes to the test stub in tests/unit/bridge/methods.test.ts.

…o-skill prompts Three fixes targeting noise reduction and correctness in the Hermes adapter pipeline, plus two Copilot review follow-ups: 1. Remove adapter-initiated episode pre-creation (__init__.py:165). The orchestrator already creates episodes on first real turn via openEpisodeIfNeeded(), so the pre-created empty episode was always abandoned as a 0-turn orphan. 2. Finalize stale episodes with content instead of abandoning them. autoFinalizeStaleTasks() now checks traceIds: episodes with at least one trace are finalized (triggering reflection + scoring), while truly empty episodes are still abandoned. 3. Strip Hermes auto-skill evaluation prompts in _turn_end(). Hermes injects a structured "review the conversation and decide whether to save a skill" prompt at task end. The adapter now detects the distinctive header phrase and truncates the system-generated scaffolding. 4. Add macOS ps(1) fallback for _is_bridge_process() (daemon_manager.py). /proc is not available on macOS, so PID validation always returned False. Fall back to `ps -p <pid> -o command=` when /proc is missing. 5. Add episode.delete / episode.delete_bulk JSON-RPC methods. HTTP routes had DELETE endpoints but the bridge lacked corresponding RPC methods, creating API surface inconsistency.

Starfie1d1272 · 2026-04-29T13:09:29Z

Copilot Review — Resolution Summary

All 13 Copilot comments have been reviewed and addressed (full replies in each thread).

Fixed (6)

#	Issue	Fix
5	macOS `/proc` not available	Added `ps -p <pid> -o command=` fallback
6	JSON-RPC missing delete methods	Added `episode.delete` + `episode.delete_bulk` with validation
7	`PermissionError` treated as dead PID	Split into own handler — PermissionError = alive
8	`ps` fallback matches too broadly	Tightened to `bridge.cts` + returncode check
9	`delete_bulk` no element validation	Per-element non-empty-string check
15	Test stub missing new methods	Added `deleteEpisode`/`deleteEpisodes` to stub

False positives (3)

#	Issue	Why
2	`_episode_id` empty → invalid turn.end	Orchestrator resolves via `openEpisodeBySession` cache (line 655)
3	`episode.close` with empty id	Wrapped in `suppress(Exception)`, pipeline handles finalize
14	PR description mismatch	Description updated — filter is in Hermes adapter as intended

Deferred — separate PRs (4)

#	Issue	Reason
1	TCP broadcast backpressure	Real but low-impact (few clients); needs TCP hardening PR
10	TCP startup failure not fatal in daemon mode	Design decision; needs dedicated daemon PR
12	`deleteEpisodes` not atomic	Needs transaction wrapping; bulk safety PR
13	TCP buffer no size limit	DoS hardening; part of TCP hardening PR

From prior PR #1546 (now closed) — requires architectural work

#	Issue
4, 11	Orphan episode hydration on restart (duplicate)

🤖 Generated with Claude Code

…_content for auto-skill eval - abandon(): when rTask is set, mark episode as finalized instead of abandoned, since the reward pipeline already completed. Log at info level and skip episode.abandoned event. - _turn_end(): apply _AUTO_SKILL_EVAL_RE to user_content too, so the review prompt sent as user_message to the fork agent is stripped.

… turn.start - reward.ts: stamp closeReason="finalized" when async scoring completes, so episodes abandoned before scoring (rTask was NULL at closeSession) get corrected to "finalized" once the reward pipeline finishes. - memos_provider/__init__.py: check query against _AUTO_SKILL_EVAL_RE in _run() before _turn_start(), so review prompts never create traces. - memory-core.ts: add TODO for orphan close bypassing capture+reward chain.

Starfie1d1272 and others added 18 commits April 27, 2026 13:29

fix(memos-local-plugin): read bridge version from package.json instea…

27d64eb

…d of hardcoding Previously the bridge reported '2.0.0-alpha.1' regardless of the actual package version. Now it reads from package.json at startup.

Update apps/memos-local-plugin/agent-contract/memory-core.ts

f2571e7

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update apps/memos-local-plugin/core/session/manager.ts

68bfe16

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

fix: reconnect FD leak, tcp close race, deleteEpisode non-optional

e09f116

bridge_client.py: close old transport before reconnect bridge/tcp.ts: await server.close() in TcpServerHandle.close() agent-contract/memory-core.ts: deleteEpisode/deleteEpisodes no longer optional

fix: shutdown re-entrancy, JSON-RPC method validation, socket error c…

7e08cf7

…leanup bridge.cts: add shuttingDown guard to prevent double-shutdown bridge/tcp.ts: validate typeof msg.method === 'string'; sock.destroy() on error to prevent resource leak

fix: remove duplicate tcpServer declaration

981a1db

fix: socket destroy guard, deleteClosedEpisode helper

a641b1d

bridge/tcp.ts: check sock.destroyed before destroying core/pipeline/memory-core.ts: extract deleteClosedEpisode helper with no-op regression guard

test(memos-local-plugin): add orphan init and deleteEpisode unit tests

8abb30b

fix(memos-local-plugin): close _rfile leak, PID path via MEMOS_HOME, …

700f006

…stale PID validation, daemon stdio gate

chore(memos-local-plugin): sync package-lock.json to 2.0.0-beta.1

3fa7375

Copilot AI review requested due to automatic review settings April 29, 2026 09:23

Copilot started reviewing on behalf of Starfie1d1272 April 29, 2026 09:24 View session

Copilot AI reviewed Apr 29, 2026

View reviewed changes

Starfie1d1272 force-pushed the fix/hydrogen-noise-and-idle-timeout branch from 1dae053 to c088cdb Compare April 29, 2026 09:40

Starfie1d1272 mentioned this pull request Apr 29, 2026

fix(memos-local-plugin): prevent orphan episode scan from closing active sessions; add episode delete API #1546

Closed

4 tasks

Starfie1d1272 force-pushed the fix/hydrogen-noise-and-idle-timeout branch 2 times, most recently from f6ebf9d to ebc5788 Compare April 29, 2026 11:48

Starfie1d1272 requested a review from Copilot April 29, 2026 11:52

Copilot started reviewing on behalf of Starfie1d1272 April 29, 2026 11:53 View session

Copilot AI reviewed Apr 29, 2026

View reviewed changes

Starfie1d1272 force-pushed the fix/hydrogen-noise-and-idle-timeout branch from ebc5788 to f191b15 Compare April 29, 2026 13:05

Starfie1d1272 added 3 commits April 30, 2026 14:03

chore: remove TODO for orphan scan — deliberate design, not a bug

8d3ed61

Conversation

Starfie1d1272 commented Apr 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

1. Remove adapter-initiated episode pre-creation

2. Finalize idle-timeout episodes instead of abandoning them

3. Filter Hermes/OpenClaw auto-skill evaluation prompts

Test plan

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Starfie1d1272 commented Apr 29, 2026

Copilot Review — Resolution Summary

Fixed (6)

False positives (3)

Deferred — separate PRs (4)

From prior PR #1546 (now closed) — requires architectural work

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Starfie1d1272 commented Apr 29, 2026 •

edited

Loading