fix(telegram): last draft push is the final text so the client folds it into the message

src/beaver_gateway/frontends/telegram/drafts.py

@@ -1,5 +1,9 @@
 """One ``sendMessageDraft`` stream per running turn (§3.8).
 
+The client folds a draft into the message that follows only when their
+texts are identical - so the last push is the final text itself, rendered
+exactly as the outbox will send it, with no status line.
+
 A draft is ephemeral and lives 30 s, Telegram throttles edits to about one
 per second per chat, and thinking or a tool call would otherwise look like a
 hang - so the draft opens with a status line straight away, is refreshed on
@@ -74,15 +78,29 @@ class Draft:
         The final ``sendMessage`` must reach Telegram after the last draft,
         or the draft lands on top of the message and lingers for its 30 s.
         """
-        if self._task is None:
-            return
+        if self._task is not None:
             self._task.cancel()
-        with contextlib.suppress(asyncio.CancelledError):
-            await self._task
+            await asyncio.gather(self._task, return_exceptions=True)
             self._task = None
         if self._inflight is not None:
             with contextlib.suppress(Exception):
                 await self._inflight
+            self._inflight = None
+
+    async def finish(self, text: str) -> None:
+        """Last push: the final text verbatim, so the message replaces the draft."""
+        await self.stop()
+        if self._broken or not text.strip():
+            return
+        try:
+            try:
+                await self._send(to_html(text), "HTML")
+            except TelegramBadRequest:
+                await self._send(text, None)
+        except TelegramAPIError as exc:
+            _log.warning(
+                "final draft to %s/%s failed: %s", self._chat_id, self._thread_id, exc
+            )
 
     async def _run(self) -> None:
         while not self._broken:

src/beaver_gateway/frontends/telegram/frontend.py

@@ -39,7 +39,7 @@ from beaver_gateway.frontends.base import Frontend
 from beaver_gateway.frontends.telegram.drafts import Draft
 from beaver_gateway.frontends.telegram.inbox import Inbox
 from beaver_gateway.frontends.telegram.outbox import Outbox
-from beaver_gateway.frontends.telegram.render import status_label
+from beaver_gateway.frontends.telegram.render import chunks, status_label
 
 if TYPE_CHECKING:
     from beaver_gateway.core.bus import Event, EventBus
@@ -659,7 +659,9 @@ class TelegramFrontend(Frontend):
                         f"⏳ {status_label(str(event['name']), event.get('input'))}"
                     )
             case "turn.end":
-                await self._close_draft(key)
+                draft = self._drafts.get(key)
+                if draft is not None:
+                    await draft.stop()
                 if event.get("stop") == "error" and event.get("origin") == "user":
                     await self._deliver(
                         conv,
@@ -691,7 +693,11 @@ class TelegramFrontend(Frontend):
     async def _on_reply(self, conv: Conversation, event: Event) -> None:
         turn_id = str(event.get("turn_id") or "")
         origin = str(event.get("item_origin") or "")
+        text = str(event.get("text") or "")
         await self._unreact(event.get("item"))
+        draft = self._drafts.pop(conv.external_id, None)
+        if draft is not None:
+            await draft.finish(chunks(text)[-1] if text.strip() else "")
         if origin != FRONTEND and not origin.startswith("сид"):
             user_text = str(event.get("user_text") or "")
             if user_text:
@@ -701,9 +707,7 @@ class TelegramFrontend(Frontend):
                     turn_id=turn_id,
                     key=f"{turn_id}:mirror",
                 )
-        await self._deliver(
-            conv, str(event.get("text") or ""), turn_id=turn_id, key=f"{turn_id}:reply"
-        )
+        await self._deliver(conv, text, turn_id=turn_id, key=f"{turn_id}:reply")
 
     # ---- drafts ------------------------------------------------------------
 

tests/test_telegram.py

@@ -251,6 +251,9 @@ async def test_general_is_master_and_reply_has_no_thread(stack: Stack) -> None:
     await asyncio.sleep(0.2)
     final = max(i for i, o in enumerate(stack.bot.order) if o[0] == "message")
     assert all(o[0] != "draft" for o in stack.bot.order[final:])
+    last = stack.bot.drafts[-1]
+    assert last["text"] == reply["text"] and last["parse_mode"] == "HTML"
+    assert stack.bot.drafts[0]["text"] == "⏳ думаю"
     rows = await stack.world.conversations.queue.recent(master.id)
     assert [r.origin for r in rows] == ["telegram"]