feat: implement conversational AI agent with tool-calling

Prototype implementation that allows:
1. Normal conversation with the AI
2. Tool-calling to update trading strategies

Created new ConversationalAgent that uses CrewAI with tools:
- get_current_strategy: Check current bot strategy
- update_trading_strategy: Update bot's trading configuration

The agent can now respond to questions like 'What is this?' without
forcing JSON output, and can update strategies when user provides
specific parameters.

Refs #51

src/backend/app/api/bots.py

@@ -16,6 +16,7 @@ from ..db.schemas import (
 )
 from ..db.models import Bot, BotConversation, User
 from ..services.ai_agent.crew import get_trading_crew
+from ..services.ai_agent.conversational import get_conversational_agent
 
 router = APIRouter()
 MAX_BOTS_PER_USER = 3
@@ -183,21 +184,20 @@ def chat(
         .order_by(BotConversation.created_at)
         .all()
     )
-    history_for_crew = [
+    history_for_agent = [
         {"role": conv.role, "content": conv.content}
         for conv in conversation_history[-10:]
     ]
 
     user_message = request.message
-    if request.strategy_config:
-        crew = get_trading_crew()
-        result = crew.chat(user_message, history_for_crew)
+    
+    # Use ConversationalAgent for natural chat with tool-calling
+    agent = get_conversational_agent(bot_id=bot_id)
+    result = agent.chat(user_message, history_for_agent)
 
     assistant_content = result.get("response", "I couldn't process your request.")
-        if result.get("success") and result.get("strategy_config"):
-            bot.strategy_config = result["strategy_config"]
-            db.commit()
 
+    # Save conversation
     db_conversation = BotConversation(
         bot_id=bot_id,
         role="user",
@@ -214,36 +214,13 @@ def chat(
     db.commit()
     db.refresh(db_assistant)
 
-        return BotChatResponse(
-            response=assistant_content,
-            strategy_config=result.get("strategy_config"),
-            success=result.get("success", False),
-        )
-    else:
-        crew = get_trading_crew()
-        result = crew.chat(user_message, history_for_crew)
-
-        assistant_content = result.get("response", "I couldn't process your request.")
-
-        db_conversation = BotConversation(
-            bot_id=bot_id,
-            role="user",
-            content=user_message,
-        )
-        db.add(db_conversation)
-
-        db_assistant = BotConversation(
-            bot_id=bot_id,
-            role="assistant",
-            content=assistant_content,
-        )
-        db.add(db_assistant)
-        db.commit()
-        db.refresh(db_assistant)
+    # If strategy was updated via tool, refresh bot data
+    if result.get("strategy_updated"):
+        db.refresh(bot)
 
     return BotChatResponse(
         response=assistant_content,
-            strategy_config=result.get("strategy_config"),
+        strategy_config=bot.strategy_config if result.get("strategy_updated") else None,
         success=result.get("success", False),
     )
 

src/backend/app/services/ai_agent/conversational.py

@@ -0,0 +1,167 @@
+"""
+Conversational Trading Agent
+
+This agent can:
+1. Have normal conversations with users
+2. Update trading strategies when user provides specific instructions
+
+Uses CrewAI's tool-calling capabilities for structured updates.
+"""
+
+from typing import List, Optional, Dict, Any
+from crewai import Agent, LLM
+from crewai.tools import tool
+from sqlalchemy.orm import Session
+
+from ...core.config import get_settings
+from ...db.models import Bot
+
+
+# Tool definitions
+@tool
+def get_current_strategy(bot_id: str) -> str:
+    """Get the current trading strategy configuration for a bot.
+    
+    Use this tool to check the current strategy before making changes.
+    
+    Args:
+        bot_id: The ID of the bot to get strategy for
+        
+    Returns:
+        JSON string with current strategy configuration
+    """
+    from ...core.database import get_db
+    from ...db.models import Bot
+    
+    db = next(get_db())
+    bot = db.query(Bot).filter(Bot.id == bot_id).first()
+    if not bot:
+        return '{"error": "Bot not found"}'
+    return str(bot.strategy_config)
+
+
+@tool
+def update_trading_strategy(
+    bot_id: str,
+    conditions: List[Dict],
+    actions: List[Dict],
+    risk_management: Optional[Dict] = None
+) -> str:
+    """Update the trading strategy configuration for a bot.
+    
+    Call this tool when the user provides specific trading parameters like:
+    - Buy/sell conditions (price drops, price rises, etc.)
+    - Take profit percentages
+    - Stop loss percentages
+    
+    Args:
+        bot_id: The ID of the bot to update
+        conditions: List of trigger conditions (e.g., [{"type": "price_drop", "token": "PEPE", "threshold": 5}])
+        actions: List of actions to take (e.g., [{"type": "buy", "amount_percent": 50}])
+        risk_management: Optional risk settings (e.g., {"stop_loss_percent": 10, "take_profit_percent": 50})
+        
+    Returns:
+        Confirmation message with updated strategy
+    """
+    from ...core.database import get_db
+    from ...db.models import Bot
+    
+    db = next(get_db())
+    bot = db.query(Bot).filter(Bot.id == bot_id).first()
+    if not bot:
+        return '{"error": "Bot not found"}'
+    
+    new_config = {
+        "conditions": conditions,
+        "actions": actions,
+    }
+    if risk_management:
+        new_config["risk_management"] = risk_management
+    
+    bot.strategy_config = new_config
+    db.commit()
+    
+    return f'Successfully updated trading strategy. New config: {new_config}'
+
+
+SYSTEM_PROMPT = """You are a helpful AI trading assistant. You can:
+
+1. Have normal conversations - answer questions about trading, tokens, strategies, etc.
+2. Help users configure their trading bots when they provide specific parameters
+
+When a user asks general questions, just answer conversationally.
+When a user provides specific trading parameters (like percentages, tokens, conditions), 
+use the update_trading_strategy tool to save their configuration.
+
+Example conversations:
+- User: "What is this?" → Answer conversationally about the trading bot platform
+- User: "I want take profit at 200%" → Use update_trading_strategy with that parameter
+- User: "Alert me when PEPE drops 5%" → Use update_trading_strategy with that condition
+
+Be friendly, helpful, and clear in your responses."""
+
+
+class ConversationalAgent:
+    def __init__(self, api_key: str, model: str = "MiniMax-M2.7", bot_id: str = None):
+        self.api_key = api_key
+        self.model = model
+        self.bot_id = bot_id
+        self.llm = LLM(
+            model=model,
+            api_key=api_key,
+            api_base="https://api.minimax.io/v1"
+        )
+        
+        # Create agent with tools
+        self.agent = Agent(
+            role="Trading Assistant",
+            goal="Help users with trading strategies and general questions",
+            backstory=SYSTEM_PROMPT,
+            tools=[get_current_strategy, update_trading_strategy],
+            llm=self.llm,
+            verbose=True,
+            allow_delegation=False,
+        )
+    
+    def chat(self, user_message: str, conversation_history: List[Dict] = None) -> Dict[str, Any]:
+        """Process a user message and return a response.
+        
+        Args:
+            user_message: The user's message
+            conversation_history: Optional list of previous messages
+            
+        Returns:
+            Dict with 'response' (the assistant's reply) and 'strategy_updated' (bool)
+        """
+        # Execute agent using kickoff
+        try:
+            result = self.agent.kickoff(user_message)
+            
+            # Check if strategy was updated
+            result_str = str(result)
+            strategy_updated = "update_trading_strategy" in result_str or \
+                             "Successfully updated" in result_str
+            
+            return {
+                "response": result_str,
+                "strategy_updated": strategy_updated,
+                "success": True
+            }
+        except Exception as e:
+            return {
+                "response": f"I encountered an error: {str(e)}. Please try again.",
+                "strategy_updated": False,
+                "success": False
+            }
+
+
+def get_conversational_agent(api_key: str = None, model: str = None, bot_id: str = None) -> ConversationalAgent:
+    """Get or create a ConversationalAgent instance."""
+    if api_key is None:
+        settings = get_settings()
+        api_key = settings.MINIMAX_API_KEY
+    if model is None:
+        settings = get_settings()
+        model = settings.MINIMAX_MODEL
+    
+    return ConversationalAgent(api_key=api_key, model=model, bot_id=bot_id)

src/backend/app/services/ai_agent/llm_connector.py

@@ -33,7 +33,7 @@ class MiniMaxLLM:
 
 
 class MiniMaxConnector:
-    def __init__(self, api_key: str, model: str = "MiniMax-Text-01"):
+    def __init__(self, api_key: str, model: str = "MiniMax-M2.7"):
         self.api_key = api_key
         self.model = model
 

src/frontend/src/lib/api/client.ts

@@ -104,11 +104,12 @@ export const api = {
 			}
 		},
 
-		async chat(id: string, message: string): Promise<BotChatResponse> {
+		async chat(id: string, message: string, signal?: AbortSignal): Promise<BotChatResponse> {
 			const response = await fetch(`${API_URL}/bots/${id}/chat`, {
 				method: 'POST',
 				headers: getAuthHeaders(),
-				body: JSON.stringify({ message } as BotChatRequest)
+				body: JSON.stringify({ message } as BotChatRequest),
+				signal
 			});
 			return handleResponse<BotChatResponse>(response);
 		},

src/frontend/src/lib/stores/chatStore.ts

@@ -8,12 +8,25 @@ export interface ChatMessage {
 	timestamp: Date;
 }
 
+// Fallback UUID generator for environments where crypto.randomUUID is not available
+function generateId(): string {
+	if (typeof crypto !== 'undefined' && typeof crypto.randomUUID === 'function') {
+		return crypto.randomUUID();
+	}
+	// Fallback: simple UUID v4 implementation
+	return 'xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx'.replace(/[xy]/g, (c) => {
+		const r = (Math.random() * 16) | 0;
+		const v = c === 'x' ? r : (r & 0x3) | 0x8;
+		return v.toString(16);
+	});
+}
+
 export const chatStore = writable<ChatMessage[]>([]);
 
 export function addMessage(message: Omit<ChatMessage, 'id' | 'timestamp'>) {
 	const newMessage: ChatMessage = {
 		...message,
-		id: crypto.randomUUID(),
+		id: generateId(),
 		timestamp: new Date()
 	};
 	chatStore.update(messages => [...messages, newMessage]);

src/frontend/src/routes/bot/[id]/+page.svelte

@@ -44,8 +44,17 @@
 		
 		isSending = true;
 
+		// Add user's message immediately so it shows even before API response
+		addMessage({ role: 'user', content: message });
+
 		try {
-			const response = await api.bots.chat(botId, message);
+			// Add timeout to prevent hanging requests
+			const controller = new AbortController();
+			const timeoutId = setTimeout(() => controller.abort(), 30000);
+			
+			const response = await api.bots.chat(botId, message, controller.signal);
+			clearTimeout(timeoutId);
+			
 			addMessage({ role: 'assistant', content: response.response });
 			
 			if (response.strategy_config) {
@@ -53,7 +62,11 @@
 				setCurrentBot(bot);
 			}
 		} catch (e) {
+			if (e instanceof Error && e.name === 'AbortError') {
+				addMessage({ role: 'assistant', content: 'Request timed out. Please try again.' });
+			} else {
 				addMessage({ role: 'assistant', content: 'Sorry, I encountered an error. Please try again.' });
+			}
 		} finally {
 			isSending = false;
 		}