core/agent: derive pending_retry_text from trailing user message

Drop stored pending field; incomplete turns are just history ending in
UserChatMessage. Compact seeds use an assistant summary so they are not
mistaken for retryable orphans.

src/plyngent/agent/chat.py

@@ -43,7 +43,6 @@ class ChatAgent:
     parallel_tools: bool
     max_context_tokens: int
     messages: list[AnyChatMessage]
-    pending_retry_text: str | None
     session_usage: TokenUsage
     last_turn_usage: TokenUsage
     last_request_usage: TokenUsage
@@ -81,13 +80,22 @@ class ChatAgent:
         self.parallel_tools = parallel_tools
         self.max_context_tokens = max_context_tokens
         self.messages = list(messages) if messages is not None else []
-        self.pending_retry_text = None
         self.session_usage = TokenUsage()
         self.last_turn_usage = TokenUsage()
         self.last_request_usage = TokenUsage()
         self.last_turn_rounds = 0
         self._ensure_system_prompt()
-        self._sync_pending_from_orphan_user()
+
+    @property
+    def pending_retry_text(self) -> str | None:
+        """Text of an incomplete last user turn, if any.
+
+        Derived only: history ending with a user message means that turn never
+        completed (failure, cancel, or resume of an orphan user in DB).
+        """
+        if self.messages and isinstance(self.messages[-1], UserChatMessage):
+            return self.messages[-1].content
+        return None
 
     @property
     def context_tokens(self) -> int:
@@ -116,13 +124,6 @@ class ChatAgent:
             return
         self.messages.insert(0, SystemChatMessage(content=self.system_prompt))
 
-    def _sync_pending_from_orphan_user(self) -> None:
-        """If history ends with a user message, that turn is incomplete → retryable."""
-        if self.messages and isinstance(self.messages[-1], UserChatMessage):
-            self.pending_retry_text = self.messages[-1].content
-        else:
-            self.pending_retry_text = None
-
     async def load_history(self) -> None:
         """Replace in-memory messages from the bound memory session."""
         if self.memory is None or self.session_id is None:
@@ -130,7 +131,6 @@ class ChatAgent:
             raise RuntimeError(msg)
         self.messages = await self.memory.list_messages(self.session_id)
         self._ensure_system_prompt()
-        self._sync_pending_from_orphan_user()
 
     async def bind_session(self, session_id: int, *, load: bool = True) -> None:
         """Attach a memory session id; optionally load existing messages."""
@@ -149,7 +149,6 @@ class ChatAgent:
         for i in range(len(self.messages) - 1, -1, -1):
             if self.messages[i] is user_msg:
                 return i
-        # Fallback: last matching content user message
         for i in range(len(self.messages) - 1, -1, -1):
             msg = self.messages[i]
             if isinstance(msg, UserChatMessage) and msg.content == user_msg.content:
@@ -157,17 +156,16 @@ class ChatAgent:
         msg = "user message not found in history"
         raise RuntimeError(msg)
 
-    def _rollback_partial(self, user_index: int, user_text: str) -> None:
+    def _rollback_partial(self, user_index: int) -> None:
         """Drop assistant/tool messages after the user; keep user for retry/DB."""
         del self.messages[user_index + 1 :]
-        self.pending_retry_text = user_text
 
     async def _run_from_user_message(self, user_msg: UserChatMessage) -> AsyncIterator[AgentEvent]:
         """Run the tool loop for an already-appended user message.
 
         On success, persists assistant/tool messages produced after the user.
-        On failure, keeps the user message (already in memory/DB) and sets
-        ``pending_retry_text`` so ``retry()`` can re-run without duplicating it.
+        On failure, keeps the trailing user message so :meth:`retry` can re-run
+        without duplicating it (see :attr:`pending_retry_text`).
         """
         user_index = self._user_index(user_msg)
 
@@ -190,7 +188,6 @@ class ChatAgent:
                 max_context_tokens=self.max_context_tokens,
             ):
                 if isinstance(event, UsageEvent):
-                    # Each tool-loop round re-sends history; sum is billing, not context size.
                     turn_rounds += 1
                     last_request = event.usage
                     turn_usage = turn_usage.add(event.usage)
@@ -199,7 +196,7 @@ class ChatAgent:
             completed = True
         except BaseException:
             if not completed:
-                self._rollback_partial(user_index, user_msg.content)
+                self._rollback_partial(user_index)
             raise
 
         self.last_turn_usage = turn_usage
@@ -207,7 +204,6 @@ class ChatAgent:
         self.last_turn_rounds = turn_rounds
         for message in self.messages[user_index + 1 :]:
             await self._persist(message)
-        self.pending_retry_text = None
 
     async def run(self, user_text: str) -> AsyncIterator[AgentEvent]:
         """Append a user message (persist immediately), run the tool loop, yield events."""
@@ -221,31 +217,13 @@ class ChatAgent:
     async def retry(self) -> AsyncIterator[AgentEvent]:
         """Re-run the incomplete last user turn without appending a new user message.
 
-        Useful after a failed/cancelled turn (user already in memory/DB) or after
-        ``load_history`` when the session ends with an orphan user message.
+        Requires history to end with a :class:`UserChatMessage` (failed/cancelled
+        turn or resumed orphan user in the session DB).
         """
-        text = self.pending_retry_text
-        if text is None:
-            self._sync_pending_from_orphan_user()
-            text = self.pending_retry_text
-        if text is None:
+        if not self.messages or not isinstance(self.messages[-1], UserChatMessage):
             msg = "nothing to retry"
             raise RuntimeError(msg)
-
+        user_msg = self.messages[-1]
         self._ensure_system_prompt()
-        if self.messages and isinstance(self.messages[-1], UserChatMessage):
-            user_msg = self.messages[-1]
-            if user_msg.content != text:
-                # Pending text out of sync: replace trailing user
-                user_msg = UserChatMessage(content=text)
-                self.messages[-1] = user_msg
-        else:
-            user_msg = UserChatMessage(content=text)
-            self.messages.append(user_msg)
-            # Already persisted on the original run when memory is bound; only
-            # persist if this is a reconstructed orphan without DB (no memory).
-            if self.memory is None or self.session_id is None:
-                await self._persist(user_msg)
-
         async for event in self._run_from_user_message(user_msg):
             yield event

src/plyngent/agent/compact.py

@@ -129,7 +129,11 @@ def build_compacted_seed_messages(
     system_prompt: str | None = None,
     source_session_id: int | None = None,
 ) -> list[AnyChatMessage]:
-    """Messages to seed a new session after compact."""
+    """Messages to seed a new session after compact.
+
+    Summary is an assistant message so history does not end with a user turn
+    (which would look like an incomplete /retry-able request).
+    """
     out: list[AnyChatMessage] = []
     if system_prompt:
         out.append(SystemChatMessage(content=system_prompt))
@@ -139,5 +143,5 @@ def build_compacted_seed_messages(
         f"{summary}\n\n"
         "Continue from this summary. Prefer not to re-ask for information already covered."
     )
-    out.append(UserChatMessage(content=body))
+    out.append(AssistantChatMessage(content=body))
     return out

src/plyngent/cli/repl.py

@@ -210,7 +210,6 @@ def _cmd_rounds(state: ReplState, arg: str) -> None:
 
 def _cmd_clear(state: ReplState) -> None:
     state.agent.messages.clear()
-    state.agent.pending_retry_text = None
     click.echo("conversation cleared (in-memory only; DB history kept)")
 
 

src/plyngent/cli/retry.py

@@ -140,9 +140,8 @@ async def run_turn_with_retries(
     """Run a chat turn with automatic retries on failure.
 
     ``starter`` produces the event stream for the first attempt (usually
-    ``agent.run``). After a failure that sets ``pending_retry_text``, further
-    attempts use ``agent.retry`` so the user message is not duplicated in
-    history/DB (it is already persisted).
+    ``agent.run``). After a failure (history ends with the user message),
+    further attempts use ``agent.retry`` so the user message is not duplicated.
 
     Ctrl+C cancels the in-flight task; user message stays in DB for ``/retry``.
     """

src/plyngent/cli/state.py

@@ -107,7 +107,6 @@ class ReplState:
         session = await self.memory.create_session(name=name, workspace=self.workspace)
         self.session_id = session.sid
         self.agent = self._make_agent()
-        self.agent.pending_retry_text = None
 
     async def resume_session(self, session_id: int) -> None:
         """Load a session; on workspace mismatch, prompt keep / rebind / abort."""
@@ -202,5 +201,4 @@ class ReplState:
         self.agent.messages = list(seed)
         for message in seed:
             _ = await self.memory.append_message(new_id, message)
-        self.agent.pending_retry_text = None
         return old_id, new_id, summary