emdash-cli 0.1.46__py3-none-any.whl → 0.1.70__py3-none-any.whl

emdash_cli/integrations/telegram/bridge.py

@@ -0,0 +1,980 @@
+"""Telegram-EmDash bridge for SSE streaming.
+
+Connects Telegram messages to the EmDash agent and streams
+SSE responses back to Telegram as message updates.
+"""
+
+import asyncio
+import json
+import os
+import time
+from dataclasses import dataclass, field
+from typing import Any, AsyncIterator
+
+import httpx
+
+from .bot import TelegramBot, TelegramMessage, TelegramUpdate
+from .config import TelegramConfig, TelegramSettings
+from .formatter import SSEEventFormatter, TelegramMessage as FormattedMessage
+
+
+# Default EmDash server URL
+DEFAULT_SERVER_URL = "http://localhost:8765"
+
+# Telegram-only commands (not forwarded to agent)
+# All other /commands are forwarded to the EmDash agent
+TELEGRAM_COMMANDS = {
+    "/start",      # Telegram welcome message
+    "/stop",       # Cancel current operation
+    "/cancel",     # Cancel pending interaction
+    "/tgstatus",   # Telegram bot status
+    "/tgsettings", # Telegram display settings
+    "/tghelp",     # Telegram help
+    "/thinking",   # Toggle thinking display
+    "/tools",      # Toggle tool calls display
+    "/plan",       # Switch to plan mode
+    "/code",       # Switch to code mode
+    "/mode",       # Show current mode
+    "/reset",      # Reset session
+}
+
+# Map BotFather command format (underscores) to EmDash format (hyphens)
+# BotFather doesn't allow hyphens in command names
+COMMAND_ALIASES = {
+    "/todo_add": "/todo-add",
+    "/verify_loop": "/verify-loop",
+}
+
+
+@dataclass
+class PendingInteraction:
+    """Tracks a pending interaction requiring user response."""
+
+    type: str  # "clarification", "plan_approval", "planmode_request"
+    data: dict = field(default_factory=dict)
+    options: list = field(default_factory=list)
+
+
+@dataclass
+class BridgeState:
+    """Tracks the state of the bridge."""
+
+    # Current session ID (one per chat)
+    sessions: dict[int, str] = field(default_factory=dict)
+
+    # Last message ID sent to each chat (for editing)
+    last_message_ids: dict[int, int] = field(default_factory=dict)
+
+    # Timestamp of last message sent to each chat (for rate limiting)
+    last_message_times: dict[int, float] = field(default_factory=dict)
+
+    # Whether we're currently processing a request for each chat
+    processing: dict[int, bool] = field(default_factory=dict)
+
+    # Pending interactions requiring user response (per chat)
+    pending: dict[int, PendingInteraction] = field(default_factory=dict)
+
+    # Current mode per chat (code or plan)
+    modes: dict[int, str] = field(default_factory=dict)
+
+
+class TelegramBridge:
+    """Bridge between Telegram and EmDash agent.
+
+    Receives messages from Telegram, sends them to the EmDash agent,
+    and streams SSE responses back as Telegram messages.
+    """
+
+    def __init__(
+        self,
+        config: TelegramConfig,
+        server_url: str | None = None,
+        on_message: Any = None,
+    ):
+        """Initialize the bridge.
+
+        Args:
+            config: Telegram configuration
+            server_url: EmDash server URL (default: localhost:8765)
+            on_message: Optional callback for logging/debugging
+        """
+        self.config = config
+        self.server_url = server_url or os.getenv("EMDASH_SERVER_URL", DEFAULT_SERVER_URL)
+        self.on_message = on_message
+
+        self.state = BridgeState()
+        self._running = False
+        self._bot: TelegramBot | None = None
+        self._http_client: httpx.AsyncClient | None = None
+
+    async def start(self) -> None:
+        """Start the bridge.
+
+        Begins listening for Telegram messages and processing them.
+        """
+        if not self.config.bot_token:
+            raise ValueError("Bot token not configured")
+
+        self._running = True
+        self._bot = TelegramBot(self.config.bot_token)
+        self._http_client = httpx.AsyncClient(timeout=None)
+
+        await self._bot.__aenter__()
+
+        # Get bot info
+        bot_info = await self._bot.get_me()
+        if self.on_message:
+            self.on_message("bridge_started", {"bot": bot_info.username})
+
+        # Start polling loop
+        try:
+            async for update in self._bot.poll_updates(
+                offset=self.config.state.last_update_id
+            ):
+                if not self._running:
+                    break
+
+                # Update offset
+                self.config.state.last_update_id = update.update_id
+
+                # Process the update
+                await self._handle_update(update)
+
+        finally:
+            await self.stop()
+
+    async def stop(self) -> None:
+        """Stop the bridge."""
+        self._running = False
+
+        if self._bot:
+            await self._bot.__aexit__(None, None, None)
+            self._bot = None
+
+        if self._http_client:
+            await self._http_client.aclose()
+            self._http_client = None
+
+    async def _handle_update(self, update: TelegramUpdate) -> None:
+        """Handle an incoming Telegram update.
+
+        Args:
+            update: The Telegram update to process
+        """
+        if not update.message or not update.message.text:
+            return
+
+        chat_id = update.message.chat.id
+        text = update.message.text.strip()
+        user = update.message.from_user
+
+        # Log the incoming message
+        if self.on_message:
+            user_name = user.display_name if user else "Unknown"
+            self.on_message("message_received", {
+                "chat_id": chat_id,
+                "user": user_name,
+                "text": text[:100],
+            })
+
+        # Check authorization
+        if not self.config.is_chat_authorized(chat_id):
+            await self._send_message(chat_id, "Sorry, this chat is not authorized.")
+            return
+
+        # Check if already processing
+        if self.state.processing.get(chat_id):
+            await self._send_message(chat_id, "Please wait, still processing previous request...")
+            return
+
+        # Check if there's a pending interaction waiting for response
+        pending = self.state.pending.get(chat_id)
+        if pending:
+            await self._handle_pending_response(chat_id, text, pending)
+            return
+
+        # Handle special commands
+        if text.startswith("/"):
+            await self._handle_command(chat_id, text)
+            return
+
+        # Process as agent message
+        await self._process_agent_message(chat_id, text)
+
+    async def _handle_command(self, chat_id: int, text: str) -> None:
+        """Handle slash commands.
+
+        Telegram-specific commands (in TELEGRAM_COMMANDS) are handled locally.
+        All other slash commands are forwarded to the EmDash agent.
+
+        Args:
+            chat_id: Chat ID
+            text: Command text (e.g., "/plan" or "/todo_add Fix tests")
+        """
+        parts = text.split(maxsplit=1)
+        command = parts[0].lower()
+        args = parts[1] if len(parts) > 1 else ""
+
+        # Handle Telegram-specific commands locally
+        if command in TELEGRAM_COMMANDS:
+            await self._handle_telegram_command(chat_id, command, args)
+            return
+
+        # Apply command aliases (BotFather format -> EmDash format)
+        if command in COMMAND_ALIASES:
+            command = COMMAND_ALIASES[command]
+
+        # Forward to agent as a slash command
+        message = f"{command} {args}".strip() if args else command
+        await self._process_agent_message(chat_id, message)
+
+    async def _handle_telegram_command(self, chat_id: int, command: str, args: str) -> None:
+        """Handle Telegram-specific commands.
+
+        Args:
+            chat_id: Chat ID
+            command: Command name (e.g., "/start")
+            args: Command arguments
+        """
+        if command == "/start":
+            await self._send_message(
+                chat_id,
+                "*EmDash Bot*\n\n"
+                "Send me a message and I'll process it with the EmDash agent.\n\n"
+                "*Mode commands:*\n"
+                "/plan - Switch to plan mode\n"
+                "/code - Switch to code mode\n"
+                "/mode - Show current mode\n"
+                "/reset - Reset session\n\n"
+                "*Telegram-only commands:*\n"
+                "/stop - Cancel current operation\n"
+                "/cancel - Cancel pending interaction\n"
+                "/tgstatus - Show bot connection status\n"
+                "/tgsettings - Show display settings\n"
+                "/thinking - Toggle showing agent thinking\n"
+                "/tools - Toggle showing tool calls\n"
+                "/tghelp - Show this help",
+            )
+
+        elif command == "/tgstatus":
+            session_id = self.state.sessions.get(chat_id)
+            pending = self.state.pending.get(chat_id)
+            current_mode = self.state.modes.get(chat_id, "code")
+            status = "Connected" if session_id else "No active session"
+            pending_status = f"\n*Pending:* {pending.type}" if pending else ""
+            mode_emoji = "📋" if current_mode == "plan" else "💻"
+            await self._send_message(
+                chat_id,
+                f"*Status:* {status}\n"
+                f"*Mode:* {mode_emoji} {current_mode}\n"
+                f"*Server:* `{self.server_url}`"
+                f"{pending_status}",
+            )
+
+        elif command == "/stop":
+            if self.state.processing.get(chat_id):
+                self.state.processing[chat_id] = False
+                await self._send_message(chat_id, "Operation cancelled.")
+            else:
+                await self._send_message(chat_id, "No operation in progress.")
+
+        elif command == "/cancel":
+            if chat_id in self.state.pending:
+                del self.state.pending[chat_id]
+                await self._send_message(chat_id, "Pending interaction cancelled.")
+            else:
+                await self._send_message(chat_id, "No pending interaction.")
+
+        elif command == "/thinking":
+            self.config.settings.show_thinking = not self.config.settings.show_thinking
+            status = "enabled" if self.config.settings.show_thinking else "disabled"
+            await self._send_message(chat_id, f"Show thinking {status}.")
+
+        elif command == "/tools":
+            self.config.settings.show_tool_calls = not self.config.settings.show_tool_calls
+            status = "enabled" if self.config.settings.show_tool_calls else "disabled"
+            await self._send_message(chat_id, f"Show tool calls {status}.")
+
+        elif command == "/tgsettings":
+            await self._send_message(
+                chat_id,
+                "*Telegram Display Settings:*\n\n"
+                f"Show thinking: `{self.config.settings.show_thinking}`\n"
+                f"Show tools: `{self.config.settings.show_tool_calls}`\n"
+                f"Compact mode: `{self.config.settings.compact_mode}`\n"
+                f"Update interval: `{self.config.settings.update_interval_ms}ms`",
+            )
+
+        elif command == "/tghelp":
+            await self._send_message(
+                chat_id,
+                "*EmDash Telegram Bot*\n\n"
+                "Send any message or slash command to interact with the EmDash agent.\n\n"
+                "*Mode commands:*\n"
+                "/plan - Switch to plan mode (read-only exploration)\n"
+                "/code - Switch to code mode (execute changes)\n"
+                "/mode - Show current mode\n"
+                "/reset - Reset session\n\n"
+                "*Agent commands (forwarded):*\n"
+                "/todos - Show todo list\n"
+                "/status - Show project status\n"
+                "/help - Show all agent commands\n\n"
+                "*Telegram-only commands:*\n"
+                "/stop - Cancel current operation\n"
+                "/cancel - Cancel pending interaction\n"
+                "/tgstatus - Bot connection status\n"
+                "/tgsettings - Display settings\n"
+                "/thinking - Toggle thinking display\n"
+                "/tools - Toggle tool calls display\n"
+                "/tghelp - This help message\n\n"
+                "*Responding to questions:*\n"
+                "Reply with option number (1, 2, 3...) or type your answer.\n\n"
+                "*Plan approval:*\n"
+                'Reply "yes" to approve, "no" to reject, or type feedback.',
+            )
+
+        elif command == "/plan":
+            # Switch to plan mode and reset session
+            self.state.modes[chat_id] = "plan"
+            if chat_id in self.state.sessions:
+                del self.state.sessions[chat_id]
+                await self._send_message(
+                    chat_id,
+                    "✅ Switched to *plan mode* (session reset)\n\n"
+                    "_Plan mode explores the codebase and creates plans without making changes._",
+                )
+            else:
+                await self._send_message(
+                    chat_id,
+                    "✅ Switched to *plan mode*\n\n"
+                    "_Plan mode explores the codebase and creates plans without making changes._",
+                )
+
+        elif command == "/code":
+            # Switch to code mode and reset session
+            self.state.modes[chat_id] = "code"
+            if chat_id in self.state.sessions:
+                del self.state.sessions[chat_id]
+                await self._send_message(
+                    chat_id,
+                    "✅ Switched to *code mode* (session reset)\n\n"
+                    "_Code mode can execute changes and modify files._",
+                )
+            else:
+                await self._send_message(
+                    chat_id,
+                    "✅ Switched to *code mode*\n\n"
+                    "_Code mode can execute changes and modify files._",
+                )
+
+        elif command == "/mode":
+            current_mode = self.state.modes.get(chat_id, "code")
+            if current_mode == "plan":
+                await self._send_message(
+                    chat_id,
+                    "📋 Current mode: *plan*\n\n"
+                    "_Use /code to switch to code mode._",
+                )
+            else:
+                await self._send_message(
+                    chat_id,
+                    "💻 Current mode: *code*\n\n"
+                    "_Use /plan to switch to plan mode._",
+                )
+
+        elif command == "/reset":
+            # Reset session for this chat
+            if chat_id in self.state.sessions:
+                del self.state.sessions[chat_id]
+            await self._send_message(chat_id, "🔄 Session reset.")
+
+    async def _process_agent_message(self, chat_id: int, message: str) -> None:
+        """Process a message through the EmDash agent.
+
+        Args:
+            chat_id: Telegram chat ID
+            message: User message to process
+        """
+        self.state.processing[chat_id] = True
+
+        # Get or create session
+        session_id = self.state.sessions.get(chat_id)
+
+        # Create formatter with settings
+        formatter = SSEEventFormatter(
+            show_thinking=self.config.settings.show_thinking,
+            show_tools=self.config.settings.show_tool_calls,
+            compact=self.config.settings.compact_mode,
+        )
+
+        # Send typing indicator
+        if self._bot:
+            try:
+                await self._bot.send_chat_action(chat_id, "typing")
+            except Exception:
+                pass
+
+        try:
+            # Stream from agent
+            last_update_time = 0.0
+            update_interval = self.config.settings.update_interval_ms / 1000.0
+            has_sent_response = False
+
+            # Get current mode for this chat (default to code)
+            current_mode = self.state.modes.get(chat_id, "code")
+
+            async for event_type, data in self._stream_agent_chat(message, session_id, current_mode):
+                # Check if cancelled
+                if not self.state.processing.get(chat_id):
+                    break
+
+                # Update session ID if provided (don't send session_start as separate message)
+                if event_type == "session_start" and data.get("session_id"):
+                    self.state.sessions[chat_id] = data["session_id"]
+                    continue  # Skip sending session_start notification
+
+                # Skip session_end notifications (clutters the chat)
+                if event_type == "session_end":
+                    continue
+
+                # Handle interactive events - store pending state
+                if event_type == "clarification":
+                    await self._handle_clarification_event(chat_id, data)
+                    continue
+
+                if event_type == "plan_submitted":
+                    await self._handle_plan_submitted_event(chat_id, data)
+                    continue
+
+                if event_type == "plan_mode_requested":
+                    await self._handle_plan_mode_event(chat_id, data)
+                    continue
+
+                # Format the event
+                formatted = formatter.format_event(event_type, data)
+
+                # Track if we've sent a response (to avoid duplicates)
+                if event_type == "response":
+                    has_sent_response = True
+
+                # Send formatted message
+                if formatted:
+                    now = time.time()
+
+                    # Rate limit updates
+                    if formatted.is_update and (now - last_update_time) < update_interval:
+                        continue
+
+                    last_update_time = now
+
+                    # Edit or send new message
+                    if formatted.is_update and chat_id in self.state.last_message_ids:
+                        await self._edit_message(
+                            chat_id,
+                            self.state.last_message_ids[chat_id],
+                            formatted.text,
+                            formatted.parse_mode,
+                        )
+                    else:
+                        msg = await self._send_message(
+                            chat_id, formatted.text, formatted.parse_mode
+                        )
+                        if msg:
+                            self.state.last_message_ids[chat_id] = msg.message_id
+
+            # Flush any pending partial content (only if we haven't sent a full response)
+            if not has_sent_response:
+                pending_content = formatter.get_pending_content()
+                if pending_content:
+                    await self._send_message(chat_id, pending_content.text, pending_content.parse_mode)
+
+        except Exception as e:
+            if self.on_message:
+                self.on_message("error", {"error": str(e)})
+            await self._send_message(chat_id, f"Error: {str(e)}")
+
+        finally:
+            self.state.processing[chat_id] = False
+
+    async def _stream_agent_chat(
+        self,
+        message: str,
+        session_id: str | None = None,
+        mode: str = "code",
+    ) -> AsyncIterator[tuple[str, dict]]:
+        """Stream agent chat response via SSE.
+
+        Args:
+            message: User message
+            session_id: Optional session ID for continuity
+            mode: Agent mode (code or plan)
+
+        Yields:
+            Tuples of (event_type, data)
+        """
+        if not self._http_client:
+            return
+
+        payload = {
+            "message": message,
+            "options": {
+                "max_iterations": 50,
+                "verbose": True,
+                "mode": mode,
+            },
+        }
+
+        if session_id:
+            payload["session_id"] = session_id
+
+        url = f"{self.server_url}/api/agent/chat"
+
+        try:
+            async with self._http_client.stream("POST", url, json=payload) as response:
+                response.raise_for_status()
+
+                current_event = None
+
+                async for line in response.aiter_lines():
+                    line = line.strip()
+
+                    if line.startswith("event: "):
+                        current_event = line[7:]
+                    elif line.startswith("data: "):
+                        if current_event:
+                            try:
+                                data = json.loads(line[6:])
+                                if data is None:
+                                    data = {}
+                                yield current_event, data
+                            except json.JSONDecodeError:
+                                pass
+                    elif line == ": ping":
+                        # Keep-alive ping
+                        pass
+
+        except httpx.HTTPError as e:
+            yield "error", {"message": f"HTTP error: {str(e)}"}
+        except Exception as e:
+            yield "error", {"message": str(e)}
+
+    async def _send_message(
+        self,
+        chat_id: int,
+        text: str,
+        parse_mode: str | None = "Markdown",
+    ) -> TelegramMessage | None:
+        """Send a message to a chat.
+
+        Args:
+            chat_id: Target chat ID
+            text: Message text
+            parse_mode: Parse mode for formatting
+
+        Returns:
+            Sent message, or None on error
+        """
+        if not self._bot:
+            return None
+
+        try:
+            return await self._bot.send_message(
+                chat_id, text, parse_mode=parse_mode
+            )
+        except Exception as e:
+            # Try sending without parse mode if Markdown fails
+            if parse_mode:
+                try:
+                    return await self._bot.send_message(
+                        chat_id, text, parse_mode=None
+                    )
+                except Exception:
+                    pass
+
+            if self.on_message:
+                self.on_message("send_error", {"error": str(e)})
+            return None
+
+    async def _edit_message(
+        self,
+        chat_id: int,
+        message_id: int,
+        text: str,
+        parse_mode: str | None = "Markdown",
+    ) -> None:
+        """Edit an existing message.
+
+        Args:
+            chat_id: Chat containing the message
+            message_id: Message ID to edit
+            text: New text
+            parse_mode: Parse mode for formatting
+        """
+        if not self._bot:
+            return
+
+        try:
+            await self._bot.edit_message_text(
+                chat_id, message_id, text, parse_mode=parse_mode
+            )
+        except Exception:
+            # Editing can fail if message is identical - ignore
+            pass
+
+    async def _send_long_message(self, chat_id: int, text: str) -> None:
+        """Send a long message, splitting if necessary.
+
+        Args:
+            chat_id: Target chat ID
+            text: Message text (can be longer than 4096 chars)
+        """
+        max_len = self.config.settings.max_message_length
+
+        # If short enough, send as-is
+        if len(text) <= max_len:
+            await self._send_message(chat_id, text)
+            return
+
+        # Split into chunks at paragraph boundaries
+        chunks = []
+        current_chunk = ""
+
+        paragraphs = text.split("\n\n")
+        for para in paragraphs:
+            if len(current_chunk) + len(para) + 2 <= max_len:
+                if current_chunk:
+                    current_chunk += "\n\n"
+                current_chunk += para
+            else:
+                if current_chunk:
+                    chunks.append(current_chunk)
+                # If paragraph itself is too long, split it
+                if len(para) > max_len:
+                    words = para.split()
+                    current_chunk = ""
+                    for word in words:
+                        if len(current_chunk) + len(word) + 1 <= max_len:
+                            if current_chunk:
+                                current_chunk += " "
+                            current_chunk += word
+                        else:
+                            chunks.append(current_chunk)
+                            current_chunk = word
+                else:
+                    current_chunk = para
+
+        if current_chunk:
+            chunks.append(current_chunk)
+
+        # Send each chunk
+        for i, chunk in enumerate(chunks):
+            if len(chunks) > 1:
+                chunk = f"*({i + 1}/{len(chunks)})*\n\n{chunk}"
+            await self._send_message(chat_id, chunk)
+            # Small delay between chunks to avoid rate limits
+            if i < len(chunks) - 1:
+                await asyncio.sleep(0.5)
+
+    # ─────────────────────────────────────────────────────────────────────────────
+    # Interactive event handlers
+    # ─────────────────────────────────────────────────────────────────────────────
+
+    async def _handle_clarification_event(self, chat_id: int, data: dict) -> None:
+        """Handle clarification event - prompt user for response.
+
+        Args:
+            chat_id: Telegram chat ID
+            data: Clarification event data
+        """
+        question = data.get("question", "")
+        options = data.get("options", [])
+        context = data.get("context", "")
+
+        # Ensure options is a list
+        if isinstance(options, str):
+            options = [options] if options else []
+
+        # Store pending interaction
+        self.state.pending[chat_id] = PendingInteraction(
+            type="clarification",
+            data=data,
+            options=options,
+        )
+
+        # Build message
+        text = f"❓ *Question:*\n{question}"
+
+        if options:
+            text += "\n\n*Options:*"
+            for i, opt in enumerate(options, 1):
+                text += f"\n{i}. {opt}"
+            text += "\n\n_Reply with option number (1-{}) or type your answer_".format(len(options))
+        else:
+            text += "\n\n_Type your answer_"
+
+        await self._send_message(chat_id, text)
+
+    async def _handle_plan_submitted_event(self, chat_id: int, data: dict) -> None:
+        """Handle plan submitted event - prompt user for approval.
+
+        Args:
+            chat_id: Telegram chat ID
+            data: Plan submitted event data
+        """
+        plan = data.get("plan", "")
+
+        # Store pending interaction
+        self.state.pending[chat_id] = PendingInteraction(
+            type="plan_approval",
+            data=data,
+        )
+
+        # Build message
+        text = f"📋 *Plan Submitted:*\n\n{plan}"
+        text += "\n\n_Reply:_"
+        text += '\n• "approve" or "yes" to proceed'
+        text += '\n• "reject" or "no" to cancel'
+        text += "\n• Or type feedback to request changes"
+
+        await self._send_long_message(chat_id, text)
+
+    async def _handle_plan_mode_event(self, chat_id: int, data: dict) -> None:
+        """Handle plan mode request event - prompt user for approval.
+
+        Args:
+            chat_id: Telegram chat ID
+            data: Plan mode request event data
+        """
+        reason = data.get("reason", "")
+
+        # Store pending interaction
+        self.state.pending[chat_id] = PendingInteraction(
+            type="planmode_request",
+            data=data,
+        )
+
+        # Build message
+        text = "🗺️ *Plan Mode Requested*"
+        if reason:
+            text += f"\n\n{reason}"
+        text += "\n\n_Reply:_"
+        text += '\n• "approve" or "yes" to enter plan mode'
+        text += '\n• "reject" or "no" to continue without plan'
+
+        await self._send_message(chat_id, text)
+
+    async def _handle_pending_response(
+        self, chat_id: int, text: str, pending: PendingInteraction
+    ) -> None:
+        """Handle user response to a pending interaction.
+
+        Args:
+            chat_id: Telegram chat ID
+            text: User's response text
+            pending: The pending interaction
+        """
+        # Clear pending state
+        del self.state.pending[chat_id]
+
+        if pending.type == "clarification":
+            await self._process_clarification_answer(chat_id, text, pending)
+        elif pending.type == "plan_approval":
+            await self._process_plan_approval(chat_id, text, pending)
+        elif pending.type == "planmode_request":
+            await self._process_planmode_approval(chat_id, text, pending)
+
+    async def _process_clarification_answer(
+        self, chat_id: int, text: str, pending: PendingInteraction
+    ) -> None:
+        """Process clarification answer and continue agent.
+
+        Args:
+            chat_id: Telegram chat ID
+            text: User's answer
+            pending: The pending clarification
+        """
+        session_id = self.state.sessions.get(chat_id)
+        if not session_id:
+            await self._send_message(chat_id, "Error: No active session")
+            return
+
+        # Check if user replied with option number
+        answer = text.strip()
+        if pending.options and answer.isdigit():
+            idx = int(answer) - 1
+            if 0 <= idx < len(pending.options):
+                answer = pending.options[idx]
+
+        await self._send_message(chat_id, f"✅ Selected: {answer}\nContinuing...")
+
+        # Call continuation endpoint and stream response
+        await self._stream_continuation(
+            chat_id,
+            f"{self.server_url}/api/agent/chat/{session_id}/clarification/answer",
+            params={"answer": answer},
+        )
+
+    async def _process_plan_approval(
+        self, chat_id: int, text: str, pending: PendingInteraction
+    ) -> None:
+        """Process plan approval/rejection and continue agent.
+
+        Args:
+            chat_id: Telegram chat ID
+            text: User's response
+            pending: The pending plan approval
+        """
+        session_id = self.state.sessions.get(chat_id)
+        if not session_id:
+            await self._send_message(chat_id, "Error: No active session")
+            return
+
+        response = text.strip().lower()
+
+        if response in ("approve", "yes", "y", "ok", "proceed"):
+            await self._send_message(chat_id, "✅ Plan approved. Executing...")
+            await self._stream_continuation(
+                chat_id,
+                f"{self.server_url}/api/agent/chat/{session_id}/plan/approve",
+            )
+        elif response in ("reject", "no", "n", "cancel"):
+            await self._send_message(chat_id, "❌ Plan rejected.")
+            await self._stream_continuation(
+                chat_id,
+                f"{self.server_url}/api/agent/chat/{session_id}/plan/reject",
+                params={"feedback": ""},
+            )
+        else:
+            # Treat as feedback
+            await self._send_message(chat_id, f"📝 Feedback sent: {text[:50]}...")
+            await self._stream_continuation(
+                chat_id,
+                f"{self.server_url}/api/agent/chat/{session_id}/plan/reject",
+                params={"feedback": text},
+            )
+
+    async def _process_planmode_approval(
+        self, chat_id: int, text: str, pending: PendingInteraction
+    ) -> None:
+        """Process plan mode approval/rejection and continue agent.
+
+        Args:
+            chat_id: Telegram chat ID
+            text: User's response
+            pending: The pending plan mode request
+        """
+        session_id = self.state.sessions.get(chat_id)
+        if not session_id:
+            await self._send_message(chat_id, "Error: No active session")
+            return
+
+        response = text.strip().lower()
+
+        if response in ("approve", "yes", "y", "ok"):
+            await self._send_message(chat_id, "✅ Entering plan mode...")
+            await self._stream_continuation(
+                chat_id,
+                f"{self.server_url}/api/agent/chat/{session_id}/planmode/approve",
+            )
+        else:
+            await self._send_message(chat_id, "Continuing without plan mode...")
+            await self._stream_continuation(
+                chat_id,
+                f"{self.server_url}/api/agent/chat/{session_id}/planmode/reject",
+                params={"feedback": text if response not in ("reject", "no", "n") else ""},
+            )
+
+    async def _stream_continuation(
+        self,
+        chat_id: int,
+        url: str,
+        params: dict | None = None,
+    ) -> None:
+        """Stream a continuation endpoint response.
+
+        Args:
+            chat_id: Telegram chat ID
+            url: API endpoint URL
+            params: Optional query parameters
+        """
+        if not self._http_client:
+            return
+
+        self.state.processing[chat_id] = True
+
+        # Create formatter with settings
+        formatter = SSEEventFormatter(
+            show_thinking=self.config.settings.show_thinking,
+            show_tools=self.config.settings.show_tool_calls,
+            compact=self.config.settings.compact_mode,
+        )
+
+        try:
+            response_content = ""
+            last_update_time = 0.0
+            update_interval = self.config.settings.update_interval_ms / 1000.0
+
+            async with self._http_client.stream("POST", url, params=params) as response:
+                response.raise_for_status()
+
+                current_event = None
+
+                async for line in response.aiter_lines():
+                    # Check if cancelled
+                    if not self.state.processing.get(chat_id):
+                        break
+
+                    line = line.strip()
+
+                    if line.startswith("event: "):
+                        current_event = line[7:]
+                    elif line.startswith("data: "):
+                        if current_event:
+                            try:
+                                data = json.loads(line[6:])
+                                if data is None:
+                                    data = {}
+
+                                # Handle interactive events recursively
+                                if current_event == "clarification":
+                                    await self._handle_clarification_event(chat_id, data)
+                                    continue
+                                if current_event == "plan_submitted":
+                                    await self._handle_plan_submitted_event(chat_id, data)
+                                    continue
+                                if current_event == "plan_mode_requested":
+                                    await self._handle_plan_mode_event(chat_id, data)
+                                    continue
+
+                                # Format the event
+                                formatted = formatter.format_event(current_event, data)
+
+                                if current_event == "response":
+                                    response_content = data.get("content", "")
+
+                                if formatted:
+                                    now = time.time()
+                                    if formatted.is_update and (now - last_update_time) < update_interval:
+                                        continue
+                                    last_update_time = now
+
+                                    msg = await self._send_message(
+                                        chat_id, formatted.text, formatted.parse_mode
+                                    )
+                                    if msg:
+                                        self.state.last_message_ids[chat_id] = msg.message_id
+
+                            except json.JSONDecodeError:
+                                pass
+
+            # Send final response if we have one
+            if response_content and chat_id not in self.state.pending:
+                await self._send_long_message(chat_id, response_content)
+
+        except Exception as e:
+            if self.on_message:
+                self.on_message("error", {"error": str(e)})
+            await self._send_message(chat_id, f"Error: {str(e)}")
+
+        finally:
+            self.state.processing[chat_id] = False