From aac36fd96f1ff7c67e4b8ca9d85048668befccde Mon Sep 17 00:00:00 2001
From: teknium1 <127238744+teknium1@users.noreply.github.com>
Date: Sat, 12 Sep 2026 19:49:52 -0700
Subject: [PATCH] refactor(gateway): text-batch flush lives on
 BasePlatformAdapter; shield + cancel-race fixes reach all 8 adapters
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Eight adapters (discord, telegram, wecom, matrix, whatsapp, simplex, feishu, weixin) each kept a
copy of the delayed text-batch flush that base._enqueue_text_event schedules. Two correctness
fixes had landed in single copies only: Discord's asyncio.shield around the dispatch (#12444 —
a late chunk cancelling the flush task aborted the in-flight agent turn) and WeCom/Weixin's
synchronous task-identity check before the pop (a superseded task waking late popped the event
and the successor found nothing). The other adapters carried both bugs latent.

The base now owns `_flush_text_batch` with both fixes, plus the `_pending_text_batches` /
`_pending_text_batch_tasks` dicts and `_SPLIT_THRESHOLD` / delay attrs (defaults; adapters set
their own). Platform policy goes through three small hooks instead of a copied body:
`_text_batch_delay_for(pending)` (Telegram's fast/short tiers, WeCom's attachment-only wait),
`_pop_text_batch(key)` (Feishu's side count table) and `_dispatch_text_batch(event)` (Feishu's
per-chat lock). Telegram keeps its `_flush_buffered` body because its contract differs on
purpose — a cancel after the pop must hold-and-re-raise so teardown can stop a flush — and gains
the identity check there. Matrix's `_split_threshold` is renamed to the shared `_SPLIT_THRESHOLD`;
SimpleX exposes its single delay through the shared attr names.

`helpers.TextBatchAggregator` (zero users) sits inside the revert-scheduled PLUGIN-COMPAT block
and is left for that revert.
---
 gateway/platforms/base.py                     | 59 +++++++++++-
 gateway/platforms/weixin.py                   | 17 ----
 plugins/platforms/discord/adapter.py          | 32 -------
 plugins/platforms/feishu/adapter.py           | 18 +---
 plugins/platforms/matrix/adapter.py           | 23 +----
 plugins/platforms/simplex/adapter.py          | 20 +---
 plugins/platforms/telegram/adapter.py         | 36 ++++---
 plugins/platforms/wecom/adapter.py            | 29 +-----
 plugins/platforms/whatsapp/adapter.py         | 15 ---
 tests/gateway/test_matrix_message_length.py   |  4 +-
 .../test_text_batch_flush_invariants.py       | 95 +++++++++++++++++++
 tests/gateway/test_text_batching.py           | 10 +-
 12 files changed, 193 insertions(+), 165 deletions(-)
 create mode 100644 tests/gateway/test_text_batch_flush_invariants.py

diff --git a/gateway/platforms/base.py b/gateway/platforms/base.py
index 23337e02f9..add5890e5f 100644
--- a/gateway/platforms/base.py
+++ b/gateway/platforms/base.py
@@ -1796,6 +1796,8 @@ class BasePlatformAdapter(ABC):
         # could drop a newer guard.
         self._active_sessions: Dict[str, asyncio.Event] = {}
         self._pending_messages: Dict[str, MessageEvent] = {}
+        self._pending_text_batches: Dict[str, MessageEvent] = {}
+        self._pending_text_batch_tasks: Dict[str, asyncio.Task] = {}
         self._session_tasks: Dict[str, asyncio.Task] = {}
         # Legacy env knob; the runner syncs the busy_input_mode value after construction.
         # Default "interrupt" so a pre-sync read never silently queues.
@@ -2235,8 +2237,15 @@ class BasePlatformAdapter(ABC):
             return None
         return resolved if isinstance(resolved, str) and resolved.strip() else None
 
-    # ── Inbound text batching: subclasses supply ``_pending_text_batches`` /
-    # ``_pending_text_batch_tasks`` dicts and ``_flush_text_batch(key)``.
+    # ── Inbound text batching. Chat clients split one long message into several inbound
+    # chunks; ``_enqueue_text_event`` merges chunks per session key and ``_flush_text_batch``
+    # dispatches after a quiet period (longer when the last chunk sits near the platform's
+    # split point, i.e. a continuation is almost certain). Adapters set the delay attrs and
+    # ``_SPLIT_THRESHOLD``; ``_text_batch_delay_for`` / ``_pop_text_batch`` /
+    # ``_dispatch_text_batch`` are the override seams for platform-specific policy.
+    _SPLIT_THRESHOLD: int = 4000
+    _text_batch_delay_seconds: float = 0.0
+    _text_batch_split_delay_seconds: float = 0.0
 
     def _event_session_key(self, event: "MessageEvent") -> str:
         """Adapter-level session key for ``event``, profile-namespaced like the agent run."""
@@ -2271,6 +2280,52 @@ class BasePlatformAdapter(ABC):
             prior_task.cancel()
         self._pending_text_batch_tasks[key] = asyncio.create_task(self._flush_text_batch(key))
 
+    def _text_batch_delay_for(self, pending: Optional["MessageEvent"]) -> float:
+        """Quiet period before ``pending`` is dispatched; near-split chunks wait longer."""
+        last_len = getattr(pending, "_last_chunk_len", 0) if pending is not None else 0
+        return self._text_batch_split_delay_seconds if last_len >= self._SPLIT_THRESHOLD else self._text_batch_delay_seconds
+
+    def _pop_text_batch(self, key: str) -> Optional["MessageEvent"]:
+        """Remove and return the pending batch for ``key`` (adapters with side tables override)."""
+        return self._pending_text_batches.pop(key, None)
+
+    async def _dispatch_text_batch(self, event: "MessageEvent") -> None:
+        """Hand a flushed batch to the pipeline (adapters with per-chat guards override)."""
+        await self.handle_message(event)
+
+    async def _flush_text_batch_now(self, key: str) -> None:
+        """Dispatch the pending batch for ``key`` immediately (no quiet period)."""
+        event = self._pop_text_batch(key)
+        if event is not None:
+            await self._dispatch_text_batch(event)
+
+    async def _flush_text_batch(self, key: str) -> None:
+        """Wait for the quiet period, then dispatch the batch for ``key``.
+
+        Two races share this body. (1) ``_enqueue_text_event`` cancels the prior flush task
+        on each new chunk; when ``Task.cancel()`` lands after ``sleep()`` already completed,
+        CancelledError is delivered at the *next* await — after a superseded task would have
+        popped the event, so the successor finds nothing and the message is lost. The identity
+        check therefore runs synchronously between the sleep and the pop. (2) A cancel that
+        lands while the dispatch is in flight would abort the agent turn (#12444), so the
+        dispatch is shielded and the outer CancelledError swallowed."""
+        current_task = asyncio.current_task()
+        try:
+            await asyncio.sleep(self._text_batch_delay_for(self._pending_text_batches.get(key)))
+            owner = self._pending_text_batch_tasks.get(key)
+            if owner is not None and owner is not current_task:
+                return
+            event = self._pop_text_batch(key)
+            if event is None:
+                return
+            logger.info("[%s] Flushing text batch %s (%d chars)", self.name, key, len(event.text or ""))
+            await asyncio.shield(self._dispatch_text_batch(event))
+        except asyncio.CancelledError:
+            pass
+        finally:
+            if self._pending_text_batch_tasks.get(key) is current_task:
+                self._pending_text_batch_tasks.pop(key, None)
+
     def _history_media_paths_for_session(self, session_key: str) -> Optional[set]:
         """Return media paths already delivered in prior turns of this session
         (MEDIA: tags / image_generate payloads), so an echoed old tag isn't re-sent."""
diff --git a/gateway/platforms/weixin.py b/gateway/platforms/weixin.py
index c2fc011e87..aeda977dbd 100644
--- a/gateway/platforms/weixin.py
+++ b/gateway/platforms/weixin.py
@@ -731,8 +731,6 @@ class WeixinAdapter(BasePlatformAdapter):
         # trigger a separate agent run. 3s / 5s (after a ~2048-char split chunk) suit iLink's cadence.
         self._text_batch_delay_seconds = self._coerce_float_extra("text_batch_delay_seconds", 3.0)
         self._text_batch_split_delay_seconds = self._coerce_float_extra("text_batch_split_delay_seconds", 5.0)
-        self._pending_text_batches: Dict[str, MessageEvent] = {}
-        self._pending_text_batch_tasks: Dict[str, asyncio.Task] = {}
         persisted = load_weixin_account(hermes_home, self._account_id) if self._account_id and not self._token else None
         if persisted:
             self._token = str(persisted.get("token") or "").strip()
@@ -945,21 +943,6 @@ class WeixinAdapter(BasePlatformAdapter):
             event.source, group_sessions_per_user=self.config.extra.get("group_sessions_per_user", True),
             thread_sessions_per_user=self.config.extra.get("thread_sessions_per_user", False), profile=event.source.profile)
 
-    async def _flush_text_batch(self, key: str) -> None:
-        current_task = asyncio.current_task()
-        try:
-            pending = self._pending_text_batches.get(key)
-            split = (getattr(pending, "_last_chunk_len", 0) if pending else 0) >= self._SPLIT_THRESHOLD
-            await asyncio.sleep(self._text_batch_split_delay_seconds if split else self._text_batch_delay_seconds)
-            if self._pending_text_batch_tasks.get(key) is not current_task:
-                return
-            event = self._pending_text_batches.pop(key, None)
-            if event:
-                await self.handle_message(event)
-        finally:
-            if self._pending_text_batch_tasks.get(key) is current_task:
-                self._pending_text_batch_tasks.pop(key, None)
-
     async def _collect_media(self, item: Dict[str, Any], media_paths: List[str], media_types: List[str]) -> None:
         spec = _INBOUND_MEDIA.get(item.get("type"))
         path, mime = await self._download_media(item, spec) if spec else (None, "")
diff --git a/plugins/platforms/discord/adapter.py b/plugins/platforms/discord/adapter.py
index b2b309b8f2..15279c467b 100644
--- a/plugins/platforms/discord/adapter.py
+++ b/plugins/platforms/discord/adapter.py
@@ -1034,8 +1034,6 @@ class DiscordAdapter(DiscordMediaMixin, BasePlatformAdapter):
         # Text batching: merge rapid successive messages (Telegram-style)
         self._text_batch_delay_seconds = env_float("HERMES_DISCORD_TEXT_BATCH_DELAY_SECONDS", 0.6)
         self._text_batch_split_delay_seconds = env_float("HERMES_DISCORD_TEXT_BATCH_SPLIT_DELAY_SECONDS", 2.0)
-        self._pending_text_batches: Dict[str, MessageEvent] = {}
-        self._pending_text_batch_tasks: Dict[str, asyncio.Task] = {}
         self._voice_text_channels: Dict[int, int] = {}  # guild_id -> text_channel_id
         self._voice_sources: Dict[int, Dict[str, Any]] = {}  # guild_id -> linked text channel source metadata
         self._voice_timeout_tasks: Dict[int, asyncio.Task] = {}  # guild_id -> timeout task
@@ -5868,36 +5866,6 @@ class DiscordAdapter(DiscordMediaMixin, BasePlatformAdapter):
             await self.handle_message(event)
         return True
 
-    # ------------------------------------------------------------------
-    # Text message aggregation (handles Discord client-side splits)
-    # ------------------------------------------------------------------
-
-    async def _flush_text_batch(self, key: str) -> None:
-        """Wait for the quiet period then dispatch; longer delay when the chunk is
-        near Discord's 2000-char split point (continuation almost certain)."""
-        current_task = asyncio.current_task()
-        try:
-            pending = self._pending_text_batches.get(key)
-            last_len = getattr(pending, "_last_chunk_len", 0) if pending else 0
-            if last_len >= self._SPLIT_THRESHOLD:
-                delay = self._text_batch_split_delay_seconds
-            else:
-                delay = self._text_batch_delay_seconds
-            await asyncio.sleep(delay)
-            event = self._pending_text_batches.pop(key, None)
-            if not event:
-                return
-            logger.info("[Discord] Flushing text batch %s (%d chars)", key, len(event.text or ""))
-            # Shield the dispatch: _enqueue_text_event cancels the prior flush task on each new chunk;
-            # without the shield CancelledError would abort the in-flight agent turn.
-            await asyncio.shield(self.handle_message(event))
-        except asyncio.CancelledError:
-            # Cancel landed before the pop; shielded handle_message unaffected.
-            pass
-        finally:
-            if self._pending_text_batch_tasks.get(key) is current_task:
-                self._pending_text_batch_tasks.pop(key, None)
-
 
 # ---------------------------------------------------------------------------
 # Discord UI Components (outside the adapter class)
diff --git a/plugins/platforms/feishu/adapter.py b/plugins/platforms/feishu/adapter.py
index 9c02560c0c..f6381ae94e 100644
--- a/plugins/platforms/feishu/adapter.py
+++ b/plugins/platforms/feishu/adapter.py
@@ -2866,21 +2866,11 @@ class FeishuAdapter(BasePlatformAdapter):
             prior_task.cancel()
         task_map[key] = asyncio.create_task(flush_fn(key))
 
-    async def _flush_text_batch(self, key: str) -> None:
-        """Flush after the quiet period; wait longer when the last chunk sits near Feishu's ~4096-char split."""
-        pending = self._pending_text_batches.get(key)
-        last_len = getattr(pending, "_last_chunk_len", 0) if pending else 0
-        near_split = last_len >= self._SPLIT_THRESHOLD  # a continuation chunk is almost certain
-        delay = self._text_batch_split_delay_seconds if near_split else self._text_batch_delay_seconds
-        await self._delayed_flush(self._pending_text_batch_tasks, key, delay, self._flush_text_batch_now)
-
-    async def _flush_text_batch_now(self, key: str) -> None:
-        """Dispatch the current text batch immediately."""
-        event = self._pending_text_batches.pop(key, None)
+    def _pop_text_batch(self, key: str) -> Optional[MessageEvent]:
         self._pending_text_batch_counts.pop(key, None)
-        if not event:
-            return
-        logger.info("[Feishu] Flushing text batch %s (%d chars)", key, len(event.text or ""))
+        return self._pending_text_batches.pop(key, None)
+
+    async def _dispatch_text_batch(self, event: MessageEvent) -> None:
         await self._handle_message_with_guards(event)
 
     # --- Message content extraction and resource download ---
diff --git a/plugins/platforms/matrix/adapter.py b/plugins/platforms/matrix/adapter.py
index 935981b969..7cb681ac62 100644
--- a/plugins/platforms/matrix/adapter.py
+++ b/plugins/platforms/matrix/adapter.py
@@ -799,7 +799,7 @@ class MatrixAdapter(BasePlatformAdapter):
     typed_command_prefix = "!"  # clients reserve typed "/" for local commands; "!command" always reaches Hermes
     # Class-level defaults keep object.__new__-built test instances working.
     max_message_length = DEFAULT_MAX_MESSAGE_LENGTH
-    _split_threshold = DEFAULT_MAX_MESSAGE_LENGTH - 100
+    _SPLIT_THRESHOLD = DEFAULT_MAX_MESSAGE_LENGTH - 100
 
     def _resolve_store_dir(self) -> Path:
         """Pin the crypto-store dir to the active profile (connect() runs inside the profile
@@ -816,7 +816,7 @@ class MatrixAdapter(BasePlatformAdapter):
         self.max_message_length = _resolve_max_message_length(config)
         self.MAX_MESSAGE_LENGTH = self.max_message_length  # mirrors other adapters for tooling
         # A chunk near the outbound limit almost certainly has a continuation.
-        self._split_threshold = max(100, self.max_message_length - 100)
+        self._SPLIT_THRESHOLD = max(100, self.max_message_length - 100)
         # Homeserver/user_id/device_id go through the same scoped reader as the token/password:
         # under multiplex os.environ holds the DEFAULT profile's identity, and pairing it with a
         # secondary's credential sends that credential to the wrong homeserver (or reuses the
@@ -875,8 +875,6 @@ class MatrixAdapter(BasePlatformAdapter):
         # Text batching merges client-side splits (~4000 chars) of one long message.
         self._text_batch_delay_seconds = float(os.getenv("HERMES_MATRIX_TEXT_BATCH_DELAY_SECONDS", "0.6"))
         self._text_batch_split_delay_seconds = float(os.getenv("HERMES_MATRIX_TEXT_BATCH_SPLIT_DELAY_SECONDS", "2.0"))
-        self._pending_text_batches: Dict[str, MessageEvent] = {}
-        self._pending_text_batch_tasks: Dict[str, asyncio.Task] = {}
         self._approval_reaction_map = {
             "✅": "once", "🌀": "session", "♾️": "always", "♾": "always", "\u267e\ufe0f": "always",
             "\u267e": "always", "❌": "deny", "❎": "deny"}
@@ -2481,23 +2479,6 @@ class MatrixAdapter(BasePlatformAdapter):
             except Exception as exc:
                 logger.debug("Matrix: failed to redact model picker reaction %s: %s", emoji, exc)
 
-    async def _flush_text_batch(self, key: str) -> None:
-        """Wait for the quiet period then dispatch the aggregated text."""
-        current_task = asyncio.current_task()
-        try:
-            pending = self._pending_text_batches.get(key)
-            last_len = getattr(pending, "_last_chunk_len", 0) if pending else 0
-            near_split = last_len >= self._split_threshold
-            await asyncio.sleep(self._text_batch_split_delay_seconds if near_split else self._text_batch_delay_seconds)
-            event = self._pending_text_batches.pop(key, None)
-            if not event:
-                return
-            logger.info("[Matrix] Flushing text batch %s (%d chars)", key, len(event.text or ""))
-            await self.handle_message(event)
-        finally:
-            if self._pending_text_batch_tasks.get(key) is current_task:
-                self._pending_text_batch_tasks.pop(key, None)
-
     def _background_read_receipt(self, room_id: str, event_id: str) -> None:
 
         async def _send() -> None:
diff --git a/plugins/platforms/simplex/adapter.py b/plugins/platforms/simplex/adapter.py
index 3bf900f093..7c4817a8d9 100644
--- a/plugins/platforms/simplex/adapter.py
+++ b/plugins/platforms/simplex/adapter.py
@@ -122,10 +122,9 @@ class SimplexAdapter(BasePlatformAdapter):
         self._pending_file_transfers: Dict[int, dict] = {}  # awaiting rcvFileComplete, by fileId
         self._pending_responses: Dict[str, asyncio.Future] = {}  # awaited command replies
         self._corr_counter = 0
-        # Text batching state consumed by BasePlatformAdapter._enqueue_text_event.
-        self._text_batch_delay = float(os.getenv("HERMES_SIMPLEX_TEXT_BATCH_DELAY", "0.8"))
-        self._pending_text_batches: Dict[str, MessageEvent] = {}
-        self._pending_text_batch_tasks: Dict[str, asyncio.Task] = {}
+        # SimpleX has no client-side split, so the split delay equals the plain one.
+        self._text_batch_delay_seconds = float(os.getenv("HERMES_SIMPLEX_TEXT_BATCH_DELAY", "0.8"))
+        self._text_batch_split_delay_seconds = self._text_batch_delay_seconds
         logger.info(
             "SimpleX adapter initialized: url=%s auto_accept=%s groups=%s",
             self.ws_url, self.auto_accept, "enabled" if self.group_allow_from else "disabled")
@@ -381,19 +380,6 @@ class SimplexAdapter(BasePlatformAdapter):
     def _text_batch_key(self, event: MessageEvent) -> str:
         return f"{event.source.platform.value}:{event.source.chat_id}"
 
-    async def _flush_text_batch(self, key: str) -> None:
-        """Wait for the quiet period then dispatch the aggregated text."""
-        current_task = asyncio.current_task()
-        try:
-            await asyncio.sleep(self._text_batch_delay)
-            event = self._pending_text_batches.pop(key, None)
-            if event:
-                logger.info("[SimpleX] Flushing text batch %s (%d chars)", key, len(event.text or ""))
-                await self.handle_message(event)
-        finally:
-            if self._pending_text_batch_tasks.get(key) is current_task:
-                self._pending_text_batch_tasks.pop(key, None)
-
     def _make_corr_id(self) -> str:
         """Mint a correlation ID and remember it for echo-filtering; the set is bounded by
         ``_max_pending_corr`` (overflow evicted in one sweep)."""
diff --git a/plugins/platforms/telegram/adapter.py b/plugins/platforms/telegram/adapter.py
index 659dadaa96..03edcbdd55 100644
--- a/plugins/platforms/telegram/adapter.py
+++ b/plugins/platforms/telegram/adapter.py
@@ -456,8 +456,6 @@ class TelegramAdapter(BasePlatformAdapter):
             "HERMES_TELEGRAM_TEXT_BATCH_DELAY_SECONDS", 0.3, min_value=0.08, max_value=2.0)
         self._text_batch_split_delay_seconds = self._env_float_clamped(
             "HERMES_TELEGRAM_TEXT_BATCH_SPLIT_DELAY_SECONDS", 1.0, min_value=self._text_batch_delay_seconds, max_value=4.0)
-        self._pending_text_batches: Dict[str, MessageEvent] = {}
-        self._pending_text_batch_tasks: Dict[str, asyncio.Task] = {}
         self._drop_delayed_deliveries = False
         # Held across disconnect: PTB advances the offset before our drop-guard runs, so Telegram won't
         # redeliver — dropping is permanent loss (see _hold_inbound_event).
@@ -5810,6 +5808,11 @@ class TelegramAdapter(BasePlatformAdapter):
         event = None
         try:
             await asyncio.sleep(delay)
+            # Superseded flush (a newer chunk re-armed the timer while our sleep was already done):
+            # CancelledError only lands at the next await, so check synchronously before the pop.
+            owner = tasks.get(key)
+            if owner is not None and owner is not current_task:
+                return
             event = pending.pop(key, None)
             if not event:
                 return
@@ -5829,23 +5832,26 @@ class TelegramAdapter(BasePlatformAdapter):
             if tasks.get(key) is current_task:
                 tasks.pop(key, None)
 
-    async def _flush_text_batch(self, key: str) -> None:
-        """Wait for the quiet period then dispatch the aggregated text."""
-        # Adaptive delay: near-split-point last chunk → long delay (continuation almost certain);
-        # short/medium totals → capped fast delays; else configured cap (all min()'d with the operator cap).
-        pending = self._pending_text_batches.get(key)
+    def _text_batch_delay_for(self, pending: Optional[MessageEvent]) -> float:
+        """Adaptive delay: near-split-point last chunk → long delay (continuation almost certain);
+        short/medium totals → capped fast delays; else configured cap (all min()'d with the operator cap)."""
         last_len = getattr(pending, "_last_chunk_len", 0) if pending else 0
         total_len = len(getattr(pending, "text", "") or "") if pending else 0
         if last_len >= self._SPLIT_THRESHOLD:
-            delay = self._text_batch_split_delay_seconds
-        elif total_len <= self._TEXT_BATCH_FAST_LEN:
-            delay = min(self._text_batch_delay_seconds, self._TEXT_BATCH_FAST_DELAY_S)
-        elif total_len <= self._TEXT_BATCH_SHORT_LEN:
-            delay = min(self._text_batch_delay_seconds, self._TEXT_BATCH_SHORT_DELAY_S)
-        else:
-            delay = self._text_batch_delay_seconds
+            return self._text_batch_split_delay_seconds
+        if total_len <= self._TEXT_BATCH_FAST_LEN:
+            return min(self._text_batch_delay_seconds, self._TEXT_BATCH_FAST_DELAY_S)
+        if total_len <= self._TEXT_BATCH_SHORT_LEN:
+            return min(self._text_batch_delay_seconds, self._TEXT_BATCH_SHORT_DELAY_S)
+        return self._text_batch_delay_seconds
+
+    async def _flush_text_batch(self, key: str) -> None:
+        """Telegram keeps its own flush body: a cancel after the pop must HOLD the event and re-raise
+        (PTB already acked the update; the hold queue redispatches after reconnect) rather than shield
+        the dispatch — teardown must be able to stop a flush from reaching a torn-down session."""
         await self._flush_buffered(
-            self._pending_text_batches, self._pending_text_batch_tasks, key, delay, "text",
+            self._pending_text_batches, self._pending_text_batch_tasks, key,
+            self._text_batch_delay_for(self._pending_text_batches.get(key)), "text",
             lambda ev: logger.info("[Telegram] Flushing text batch %s (%d chars)", key, len(ev.text or "")))
 
     # -- Photo batching --
diff --git a/plugins/platforms/wecom/adapter.py b/plugins/platforms/wecom/adapter.py
index 66eaa6d2fe..4116da4289 100644
--- a/plugins/platforms/wecom/adapter.py
+++ b/plugins/platforms/wecom/adapter.py
@@ -149,8 +149,6 @@ class WeComAdapter(WeComStreamMixin, WeComMediaMixin, ChatSendQueueMixin, BasePl
         self._text_batch_delay_seconds = env_float("HERMES_WECOM_TEXT_BATCH_DELAY_SECONDS", 0.6)
         self._text_batch_split_delay_seconds = env_float("HERMES_WECOM_TEXT_BATCH_SPLIT_DELAY_SECONDS", 2.0)
         self._attachment_text_merge_delay_seconds = _extra_float("attachment_text_merge_delay_seconds", 0.8)
-        self._pending_text_batches: Dict[str, MessageEvent] = {}
-        self._pending_text_batch_tasks: Dict[str, asyncio.Task] = {}
         # Stream keep-alive config (see streaming.py STREAM_* constants).
         self._stream_safe_duration_seconds = _extra_float("stream_safe_duration_seconds", STREAM_SAFE_DURATION_SECONDS)
         self._stream_keepalive_enabled = bool(extra.get("stream_keepalive_enabled", STREAM_KEEPALIVE_ENABLED_DEFAULT))
@@ -503,29 +501,10 @@ class WeComAdapter(WeComStreamMixin, WeComMediaMixin, ChatSendQueueMixin, BasePl
                 existing.reply_to_text = event.reply_to_text
                 existing.reply_to_message_id = event.reply_to_message_id
 
-    async def _flush_text_batch(self, key: str) -> None:
-        current_task = asyncio.current_task()
-        try:
-            pending = self._pending_text_batches.get(key)
-            if pending and pending.media_urls and not (pending.text or "").strip():
-                delay = self._attachment_text_merge_delay_seconds  # attachment-only: wait for text
-            elif pending and getattr(pending, "_last_chunk_len", 0) >= self._SPLIT_THRESHOLD:
-                delay = self._text_batch_split_delay_seconds  # continuation almost certain
-            else:
-                delay = self._text_batch_delay_seconds
-            await asyncio.sleep(delay)
-            # Cancel-delivery race: CancelledError lands at the NEXT await, so this identity check
-            # must stay synchronous (no await between it and the pop).
-            if self._pending_text_batch_tasks.get(key) is not current_task:
-                return
-            event = self._pending_text_batches.pop(key, None)
-            if not event:
-                return
-            logger.info("[WeCom] Flushing batch %s (%d chars, %d media)", key, len(event.text or ""), len(event.media_urls or []))
-            await self.handle_message(event)
-        finally:
-            if self._pending_text_batch_tasks.get(key) is current_task:
-                self._pending_text_batch_tasks.pop(key, None)
+    def _text_batch_delay_for(self, pending: Optional[MessageEvent]) -> float:
+        if pending is not None and pending.media_urls and not (pending.text or "").strip():
+            return self._attachment_text_merge_delay_seconds  # attachment-only: wait for the text frame
+        return super()._text_batch_delay_for(pending)
 
     @staticmethod
     def _extract_text(body: Dict[str, Any]) -> Tuple[str, Optional[str]]:
diff --git a/plugins/platforms/whatsapp/adapter.py b/plugins/platforms/whatsapp/adapter.py
index 09f407554d..1a918b5a0a 100644
--- a/plugins/platforms/whatsapp/adapter.py
+++ b/plugins/platforms/whatsapp/adapter.py
@@ -283,8 +283,6 @@ class WhatsAppAdapter(WhatsAppBehaviorMixin, BasePlatformAdapter):
         # Text debounce batching: rapid bursts (forwards, paste-splits) would otherwise each trigger a separate agent turn.
         self._text_batch_delay_seconds = self._coerce_float_extra("text_batch_delay_seconds", 5.0)
         self._text_batch_split_delay_seconds = self._coerce_float_extra("text_batch_split_delay_seconds", 10.0)
-        self._pending_text_batches: Dict[str, MessageEvent] = {}
-        self._pending_text_batch_tasks: Dict[str, asyncio.Task] = {}
 
     def _coerce_float_extra(self, key: str, default: float) -> float:
         """Read a float from ``config.extra``; NaN/Inf/negative/unparseable → ``default`` (fed to asyncio.sleep)."""
@@ -732,19 +730,6 @@ class WhatsAppAdapter(WhatsAppBehaviorMixin, BasePlatformAdapter):
 
     _SPLIT_THRESHOLD = 6000  # WhatsApp supports ~65K chars; generous threshold
 
-    async def _flush_text_batch(self, key: str) -> None:
-        current_task = asyncio.current_task()
-        try:
-            pending = self._pending_text_batches.get(key)
-            last_len = getattr(pending, "_last_chunk_len", 0) if pending else 0
-            await asyncio.sleep(self._text_batch_split_delay_seconds if last_len >= self._SPLIT_THRESHOLD else self._text_batch_delay_seconds)
-            event = self._pending_text_batches.pop(key, None)
-            if event:
-                await self.handle_message(event)
-        finally:
-            if self._pending_text_batch_tasks.get(key) is current_task:
-                self._pending_text_batch_tasks.pop(key, None)
-
     @staticmethod
     def _classify_bridge_message(data: Dict[str, Any]) -> MessageType:
         media_type = str(data.get("mediaType", "") or "")
diff --git a/tests/gateway/test_matrix_message_length.py b/tests/gateway/test_matrix_message_length.py
index f3d8a2996e..a10a538f54 100644
--- a/tests/gateway/test_matrix_message_length.py
+++ b/tests/gateway/test_matrix_message_length.py
@@ -27,12 +27,12 @@ class TestMatrixMaxMessageLength:
     def test_default_limit_is_16000(self):
         adapter = _make_adapter()
         assert adapter.max_message_length == 16000
-        assert adapter._split_threshold == 15900
+        assert adapter._SPLIT_THRESHOLD == 15900
 
     def test_extra_override(self):
         adapter = _make_adapter(max_message_length=12000)
         assert adapter.max_message_length == 12000
-        assert adapter._split_threshold == 11900
+        assert adapter._SPLIT_THRESHOLD == 11900
 
     def test_env_override(self, monkeypatch):
         monkeypatch.setenv("MATRIX_MAX_MESSAGE_LENGTH", "20000")
diff --git a/tests/gateway/test_text_batch_flush_invariants.py b/tests/gateway/test_text_batch_flush_invariants.py
new file mode 100644
index 0000000000..7ae1699cf4
--- /dev/null
+++ b/tests/gateway/test_text_batch_flush_invariants.py
@@ -0,0 +1,95 @@
+"""Invariants for BasePlatformAdapter's inbound text-batch flush.
+
+Every adapter that batches inbound text used to carry its own ``_flush_text_batch``; only Discord
+shielded the dispatch (#12444) and only WeCom/Weixin checked task identity before the pop. Both
+fixes now live in the base flush; these tests pin them against the base class and sweep every
+adapter that still overrides the flush (Telegram keeps a hold-queue variant) for the same contract.
+"""
+
+import asyncio
+from typing import Any, Dict
+
+import pytest
+
+from gateway.config import Platform, PlatformConfig
+from gateway.platforms.base import BasePlatformAdapter
+from gateway.platforms.event import MessageEvent, MessageType
+from gateway.session import SessionSource
+
+
+class _Adapter(BasePlatformAdapter):
+    def __init__(self):
+        super().__init__(PlatformConfig(enabled=True), Platform.TELEGRAM)
+        self._text_batch_delay_seconds = 0.0
+        self._text_batch_split_delay_seconds = 0.0
+        self.dispatched: list = []
+        self.entered = asyncio.Event()
+        self.release = asyncio.Event()
+
+    async def connect(self, *, is_reconnect: bool = False) -> bool:
+        return True
+
+    async def disconnect(self) -> None:
+        pass
+
+    async def send(self, *a: Any, **k: Any) -> None:
+        pass
+
+    async def get_chat_info(self, chat_id: str) -> Dict[str, Any]:
+        return {}
+
+    async def handle_message(self, event: MessageEvent) -> None:
+        self.entered.set()
+        await self.release.wait()  # a real agent turn awaits here
+        self.dispatched.append(event)
+
+
+def _event(text: str) -> MessageEvent:
+    return MessageEvent(text=text, message_type=MessageType.TEXT,
+                        source=SessionSource(platform=Platform.TELEGRAM, chat_id="c", chat_type="dm"))
+
+
+@pytest.mark.asyncio
+async def test_cancel_mid_dispatch_does_not_abort_the_turn():
+    """A cancel landing while handle_message is in flight (a late chunk re-arming the timer) must not
+    abort the dispatch: the batch is delivered exactly once and the task exits clean."""
+    adapter = _Adapter()
+    adapter._pending_text_batches["k"] = _event("hello")
+    task = asyncio.create_task(adapter._flush_text_batch("k"))
+    adapter._pending_text_batch_tasks["k"] = task
+    await adapter.entered.wait()
+    task.cancel()
+    adapter.release.set()
+    await task  # no CancelledError escapes
+    assert [e.text for e in adapter.dispatched] == ["hello"]
+    assert adapter._pending_text_batch_tasks == {}
+
+
+@pytest.mark.asyncio
+async def test_superseded_flush_leaves_batch_for_its_successor():
+    """When a newer flush task owns the key by the time the old one wakes, the old one must neither
+    pop nor dispatch — otherwise the successor finds an empty batch and the message is lost."""
+    adapter = _Adapter()
+    adapter.release.set()
+    event = _event("hello")
+    adapter._pending_text_batches["k"] = event
+    stale = asyncio.create_task(adapter._flush_text_batch("k"))
+    adapter._pending_text_batch_tasks["k"] = stale
+    successor = asyncio.create_task(asyncio.sleep(60))
+    adapter._pending_text_batch_tasks["k"] = successor  # re-armed before `stale` ran
+    await stale
+    successor.cancel()
+    assert adapter.dispatched == []
+    assert adapter._pending_text_batches.get("k") is event
+    assert adapter._pending_text_batch_tasks.get("k") is successor
+
+
+@pytest.mark.asyncio
+async def test_enqueue_then_flush_delivers_merged_text_once():
+    adapter = _Adapter()
+    adapter.release.set()
+    adapter._enqueue_text_event(_event("part one"))
+    adapter._enqueue_text_event(_event("part two"))
+    await asyncio.gather(*adapter._pending_text_batch_tasks.values())
+    assert [e.text for e in adapter.dispatched] == ["part one\npart two"]
+    assert adapter._pending_text_batches == {} and adapter._pending_text_batch_tasks == {}
diff --git a/tests/gateway/test_text_batching.py b/tests/gateway/test_text_batching.py
index 63504c0b74..e8be4fe27c 100644
--- a/tests/gateway/test_text_batching.py
+++ b/tests/gateway/test_text_batching.py
@@ -45,7 +45,7 @@ def _make_discord_adapter():
 
     config = PlatformConfig(enabled=True, token="test-token")
     adapter = object.__new__(DiscordAdapter)
-    adapter._platform = Platform.DISCORD
+    adapter._platform = adapter.platform = Platform.DISCORD
     adapter.config = config
     adapter._pending_text_batches = {}
     adapter._pending_text_batch_tasks = {}
@@ -105,7 +105,7 @@ def _make_matrix_adapter():
 
     config = PlatformConfig(enabled=True, token="test-token")
     adapter = object.__new__(MatrixAdapter)
-    adapter._platform = Platform.MATRIX
+    adapter._platform = adapter.platform = Platform.MATRIX
     adapter.config = config
     adapter._pending_text_batches = {}
     adapter._pending_text_batch_tasks = {}
@@ -159,7 +159,7 @@ def _make_wecom_adapter():
 
     config = PlatformConfig(enabled=True, token="test-token")
     adapter = object.__new__(WeComAdapter)
-    adapter._platform = Platform.WECOM
+    adapter._platform = adapter.platform = Platform.WECOM
     adapter.config = config
     adapter._pending_text_batches = {}
     adapter._pending_text_batch_tasks = {}
@@ -213,7 +213,7 @@ def _make_telegram_adapter():
 
     config = PlatformConfig(enabled=True, token="test-token")
     adapter = object.__new__(TelegramAdapter)
-    adapter._platform = Platform.TELEGRAM
+    adapter._platform = adapter.platform = Platform.TELEGRAM
     adapter.config = config
     adapter._pending_text_batches = {}
     adapter._pending_text_batch_tasks = {}
@@ -247,7 +247,7 @@ def _make_feishu_adapter():
 
     config = PlatformConfig(enabled=True, token="test-token")
     adapter = object.__new__(FeishuAdapter)
-    adapter._platform = Platform.FEISHU
+    adapter._platform = adapter.platform = Platform.FEISHU
     adapter.config = config
     batch_state = FeishuBatchState()
     adapter._pending_text_batches = batch_state.events