fix: per-topic queues, 429 hardening, and interactive UI reliability

src/ccbot/bot.py

@@ -154,6 +154,7 @@
     "help": "↗ Show Claude Code help",
     "memory": "↗ Edit CLAUDE.md",
     "model": "↗ Switch AI model",
+    "context": "↗ Show context window usage",
 }
 
 
@@ -1741,12 +1742,17 @@ async def handle_new_message(msg: NewMessage, bot: Bot) -> None:
             # Mark interactive mode BEFORE sleeping so polling skips this window
             set_interactive_mode(user_id, wid, thread_id)
             # Flush pending messages (e.g. plan content) before sending interactive UI
-            queue = get_message_queue(user_id)
+            queue = get_message_queue(user_id, thread_id)
             if queue:
                 await queue.join()
-            # Wait briefly for Claude Code to render the question UI
-            await asyncio.sleep(0.3)
-            handled = await handle_interactive_ui(bot, user_id, wid, thread_id)
+            # Retry pane capture with increasing delays — Claude Code may need
+            # time to render the interactive UI after the JSONL entry is written.
+            handled = False
+            for delay in (0.3, 0.7, 1.0):
+                await asyncio.sleep(delay)
+                handled = await handle_interactive_ui(bot, user_id, wid, thread_id)
+                if handled:
+                    break
             if handled:
                 # Update user's read offset
                 session = await session_manager.resolve_session_for_window(wid)
@@ -1768,7 +1774,10 @@ async def handle_new_message(msg: NewMessage, bot: Bot) -> None:
             await clear_interactive_msg(user_id, bot, thread_id)
 
         # Skip tool call notifications when CCBOT_SHOW_TOOL_CALLS=false
-        if not config.show_tool_calls and msg.content_type in ("tool_use", "tool_result"):
+        if not config.show_tool_calls and msg.content_type in (
+            "tool_use",
+            "tool_result",
+        ):
             continue
 
         parts = build_response_parts(
@@ -1841,6 +1850,25 @@ async def post_init(application: Application) -> None:
     if rate_limiter and rate_limiter._base_limiter:
         rate_limiter._base_limiter._level = rate_limiter._base_limiter.max_rate
         logger.info("Pre-filled global rate limiter bucket")
+        # Also pre-fill per-group limiters for known chat IDs.
+        # Without this, the group limiter allows a burst of 20 requests on restart,
+        # which can exceed Telegram's persisted server-side per-group counter.
+        if hasattr(rate_limiter, "_group_limiters"):
+            from aiolimiter import AsyncLimiter
+
+            group_rate = getattr(rate_limiter, "_group_max_rate", 20)
+            group_period = getattr(rate_limiter, "_group_time_period", 60)
+            seen_chat_ids: set[int] = set()
+            for chat_id in session_manager.group_chat_ids.values():
+                if chat_id < 0 and chat_id not in seen_chat_ids:
+                    seen_chat_ids.add(chat_id)
+                    limiter = AsyncLimiter(group_rate, group_period)
+                    limiter._level = limiter.max_rate
+                    rate_limiter._group_limiters[chat_id] = limiter
+            if seen_chat_ids:
+                logger.info(
+                    "Pre-filled %d group rate limiter bucket(s)", len(seen_chat_ids)
+                )
 
     monitor = SessionMonitor()
 

src/ccbot/config.py

@@ -112,6 +112,22 @@ def __init__(self) -> None:
         for var in SENSITIVE_ENV_VARS:
             os.environ.pop(var, None)
 
+        # Display thinking messages in real-time and history
+        # Set SHOW_THINKING=false to skip thinking blocks (reduces Telegram API calls)
+        self.show_thinking = os.getenv("SHOW_THINKING", "true").lower() in (
+            "true",
+            "1",
+            "yes",
+        )
+
+        # Display tool_use/tool_result messages in real-time and history
+        # Set SHOW_TOOLS=false to skip tool messages (reduces Telegram API calls)
+        self.show_tools = os.getenv("SHOW_TOOLS", "true").lower() in (
+            "true",
+            "1",
+            "yes",
+        )
+
         logger.debug(
             "Config initialized: dir=%s, token=%s..., allowed_users=%d, "
             "tmux_session=%s, claude_projects_path=%s",

src/ccbot/handlers/history.py

@@ -160,6 +160,16 @@ async def send_history(
 
         lines = [header]
         for msg in messages:
+            # Skip thinking messages unless show_thinking is enabled
+            if msg.get("content_type") == "thinking" and not config.show_thinking:
+                continue
+            # Skip tool messages unless show_tools is enabled
+            if (
+                msg.get("content_type") in ("tool_use", "tool_result")
+                and not config.show_tools
+            ):
+                continue
+
             # Format timestamp as HH:MM
             ts = msg.get("timestamp")
             if ts: