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

802 lines
37 KiB
Python

"""
Conversational Trading Agent
This agent can:
1. Have normal conversations with users
2. Update trading strategies when user provides specific instructions
Uses MiniMax extended thinking API for proper thinking/reasoning separation.
"""
import json
import re
import requests
from typing import List, Optional, Dict, Any
from datetime import timedelta
from ...core.config import get_settings
from ...db.models import Bot
SYSTEM_PROMPT = """You are a helpful AI trading assistant named Randebu. You help users manage their trading bots.
IMPORTANT CHAIN LIMITATION:
- We ONLY support BSC (Binance Smart Chain) blockchain
- If user asks about any other chain (Solana, ETH, Base, etc.), respond with: "Currently we only support BSC (Binance Smart Chain). All trading strategies and token searches are performed on BSC."
- Never search or recommend tokens on other chains
- The search_tokens tool defaults to BSC, never change this
Your response must be valid JSON with exactly this structure:
{
"thinking": "Your internal reasoning and analysis (what you're thinking about)",
"response": "Your actual response to the user (be concise and helpful)",
"strategy_update": null or {
"conditions": [{"type": "price_drop" | "price_rise" | "volume_spike" | "price_level", "token": "TOKEN_SYMBOL", "token_address": null, "threshold": number, ...}],
"actions": [{"type": "buy" | "sell" | "hold", "amount_percent": number, ...}],
"risk_management": {"stop_loss_percent": number, "take_profit_percent": number}
}
}
Guidelines:
- "thinking" should be detailed reasoning about the user's request
- "response" should be conversational and clear
- "strategy_update" should be populated ONLY when the user provides specific trading parameters (percentages, tokens, conditions, etc.)
- IMPORTANT: When a token is mentioned, set "token_address": null and ask user to confirm the token address before saving. Your response should say something like: "I need to confirm the token address. Could you provide the contract address for [TOKEN]?"
- If no strategy parameters are provided, set "strategy_update" to null
- Be friendly, concise, and helpful in your response
Example 1 (no strategy update):
User: "What can this bot do?"
{
"thinking": "The user is asking about the bot's capabilities. I should explain the main features.",
"response": "Randebu is your AI trading assistant! It can monitor cryptocurrency prices and execute trades based on your configured strategies. Tell me your trading parameters and I'll set them up for you.",
"strategy_update": null
}
Example 2 (token needs confirmation):
User: "I want to buy PEPE when it drops 10%"
{
"thinking": "User wants to buy PEPE. I need the token contract address to proceed. I should ask for confirmation.",
"response": "I'd be happy to set up a buy order for PEPE! However, I need to confirm the token contract address. Could you provide the BSC contract address for PEPE? (It usually starts with 0x...)",
"strategy_update": {
"conditions": [{"type": "price_drop", "token": "PEPE", "token_address": null, "threshold": 10}],
"actions": [{"type": "buy", "amount_percent": 100}],
"risk_management": null
}
}
Example 3 (with token address provided by user):
User: "Buy 0x6982508145454Ce125dDE157d8d64a26D53f60a2 when it drops 10%"
{
"thinking": "User provided a contract address, I can use it directly.",
"response": "Perfect! I've configured your strategy to buy the token when it drops 10%.",
"strategy_update": {
"conditions": [{"type": "price_drop", "token": "TOKEN", "token_address": "0x6982508145454Ce125dDE157d8d64a26D53f60a2", "threshold": 10}],
"actions": [{"type": "buy", "amount_percent": 100}],
"risk_management": null
}
}"""
# Tool definitions for the agent
TOOLS = [
{
"type": "function",
"function": {
"name": "search_tokens",
"description": "Search for trending tokens on BSC blockchain. Use this when user asks for token recommendations, trending tokens, or wants to discover new tokens to trade. ALWAYS uses BSC chain.",
"parameters": {
"type": "object",
"properties": {
"limit": {
"type": "integer",
"description": "Number of tokens to return (default: 10)",
"default": 10
}
},
"required": []
}
}
},
{
"type": "function",
"function": {
"name": "run_backtest",
"description": "Run a backtest to evaluate how the current trading strategy would have performed historically. Returns key metrics like ROI, win rate, max drawdown, etc. Use this when user asks to backtest, test strategy, or check historical performance.",
"parameters": {
"type": "object",
"properties": {
"token_address": {
"type": "string",
"description": "The BSC contract address of the token to backtest (required)"
},
"timeframe": {
"type": "string",
"description": "Timeframe for klines: '1d' (1 day), '4h' (4 hours), '1h' (1 hour), '15m' (15 minutes)",
"default": "1d"
},
"start_date": {
"type": "string",
"description": "Start date for backtest in YYYY-MM-DD format (e.g., '2024-01-01')"
},
"end_date": {
"type": "string",
"description": "End date for backtest in YYYY-MM-DD format (e.g., '2024-12-01')"
}
},
"required": ["token_address"]
}
}
},
{
"type": "function",
"function": {
"name": "manage_simulation",
"description": "Manage trading simulations: start, stop, or check status. Simulations run on real-time klines and show live portfolio updates. Use when user asks to run simulation, check simulation status, or stop simulation.",
"parameters": {
"type": "object",
"properties": {
"action": {
"type": "string",
"enum": ["start", "stop", "status", "results"],
"description": "Action to perform: 'start' (begin new simulation), 'stop' (stop running simulation), 'status' (check if simulation is running), 'results' (get results from current or latest simulation)"
},
"token_address": {
"type": "string",
"description": "Token contract address for simulation (required for 'start' action)"
},
"kline_interval": {
"type": "string",
"description": "Kline interval: '1m', '5m', '15m', '1h' (default: '1m')",
"default": "1m"
}
},
"required": ["action"]
}
}
}
]
SYSTEM_PROMPT_WITH_TOOLS = SYSTEM_PROMPT + """
You have access to tools:
- search_tokens(chain, limit): Search for trending tokens on a blockchain. Use it when user asks for token recommendations or trending tokens.
- run_backtest(token_address, timeframe, start_date, end_date): Run a backtest on historical data. Returns performance metrics. Use when user asks to backtest or check historical performance.
- manage_simulation(action, token_address, kline_interval): Manage trading simulations. Actions: 'start' (begin new), 'stop' (stop running), 'status' (check if running), 'results' (get current/latest results).
When you want to use a tool, respond with:
{
"thinking": "...",
"response": "Running backtest...",
"tool_call": {"name": "run_backtest", "arguments": {"token_address": "0x...", "timeframe": "1d", "start_date": "2024-01-01", "end_date": "2024-12-01"}}
}
"""
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
# Extended thinking endpoint
self.thinking_endpoint = "https://api.minimax.io/v1/text/chatcompletion_v2"
def chat(self, user_message: str, conversation_history: List[Dict] = None) -> Dict[str, Any]:
"""Process a user message and return a structured response.
Args:
user_message: The user's message
conversation_history: Optional list of previous messages
Returns:
Dict with 'response', 'thinking', and 'strategy_updated'
"""
try:
# Build messages array with system prompt and conversation history
messages = [{"role": "system", "content": SYSTEM_PROMPT_WITH_TOOLS}]
# Add conversation history (last 10 messages)
if conversation_history:
for msg in conversation_history[-10:]:
role = "assistant" if msg.get("role") == "assistant" else "user"
messages.append({"role": role, "content": msg.get("content", "")})
# Add current user message
messages.append({"role": "user", "content": user_message})
# Make API call to extended thinking endpoint
resp = requests.post(
self.thinking_endpoint,
headers={
"Authorization": f"Bearer {self.api_key}",
"Content-Type": "application/json"
},
json={
"model": self.model,
"messages": messages,
"temperature": 0.7,
"max_tokens": 2000,
"thinking": {
"type": "human",
"budget_tokens": 1500
},
"tools": TOOLS
}
)
result = resp.json()
# Extract thinking from reasoning_content
thinking = None
if "choices" in result and len(result["choices"]) > 0:
choice = result["choices"][0]
if "message" in choice:
message = choice["message"]
thinking = message.get("reasoning_content")
# Check for native function calls (tool_calls)
tool_calls = message.get("tool_calls", [])
if tool_calls:
for tool_call in tool_calls:
func = tool_call.get("function", {})
func_name = func.get("name", "")
args = json.loads(func.get("arguments", "{}"))
if func_name == "search_tokens":
chain = "bsc" # Always BSC
limit = args.get("limit", 10)
# Execute the tool
from ..ave.client import AveCloudClient
from ...core.config import get_settings
settings = get_settings()
ave_client = AveCloudClient(
api_key=settings.AVE_API_KEY,
plan=settings.AVE_API_PLAN
)
import asyncio
tokens = asyncio.run(ave_client.get_tokens(chain=chain, limit=limit))
if tokens:
# Format tokens for response
token_list = ""
for t in tokens[:limit]:
addr = t.get("token", "")
symbol = t.get("symbol", "")
name = t.get("name", "")
price_change = t.get("token_price_change_24h", "N/A")
token_list += f"- **{symbol}** ({name}): `{addr}` - 24h change: {price_change}%\n"
response_text = f"Here are the trending tokens on {chain.upper()}:\n\n{token_list}\nWould you like me to set up a strategy for any of these?"
else:
response_text = f"I couldn't find any trending tokens on {chain.upper()}. Try again later."
# Return the tool result directly
return {
"response": response_text,
"thinking": thinking,
"strategy_updated": False,
"strategy_needs_confirmation": False,
"success": True
}
elif func_name == "run_backtest":
token_address = args.get("token_address")
timeframe = args.get("timeframe", "1d")
start_date = args.get("start_date")
end_date = args.get("end_date")
# Execute backtest
backtest_result = self._execute_backtest(
token_address=token_address,
timeframe=timeframe,
start_date=start_date,
end_date=end_date
)
return {
"response": backtest_result,
"thinking": thinking,
"strategy_updated": False,
"strategy_needs_confirmation": False,
"success": True
}
elif func_name == "manage_simulation":
action = args.get("action")
token_address = args.get("token_address")
kline_interval = args.get("kline_interval", "1m")
# Execute simulation management
sim_result = self._manage_simulation(
action=action,
token_address=token_address,
kline_interval=kline_interval
)
return {
"response": sim_result,
"thinking": thinking,
"strategy_updated": False,
"strategy_needs_confirmation": False,
"success": True
}
# Get the main response content
content = result.get("choices", [{}])[0].get("message", {}).get("content", "")
# Parse JSON from the content
thinking_field = None
response_text = content
strategy_update = None
# Try to extract JSON from the content
json_match = re.search(r'```(?:json)?\s*(\{.*?\})\s*```', content, re.DOTALL)
if json_match:
json_str = json_match.group(1)
else:
# Try to find JSON object directly
json_match = re.search(r'\{.*\}', content, re.DOTALL)
if json_match:
json_str = json_match.group(0)
else:
json_str = None
if json_str:
try:
parsed = json.loads(json_str)
thinking_field = parsed.get("thinking", "")
response_text = parsed.get("response", content)
strategy_update = parsed.get("strategy_update")
# Handle tool call
tool_call = parsed.get("tool_call")
if tool_call and tool_call.get("name") == "search_tokens":
args = tool_call.get("arguments", {})
chain = args.get("chain", "bsc")
limit = args.get("limit", 10)
# Execute the tool
from ..ave.client import AveCloudClient
from ...core.config import get_settings
settings = get_settings()
ave_client = AveCloudClient(
api_key=settings.AVE_API_KEY,
plan=settings.AVE_API_PLAN
)
import asyncio
tokens = asyncio.run(ave_client.get_tokens(chain=chain, limit=limit))
if tokens:
# Format tokens for response
token_list = ""
for t in tokens[:limit]:
addr = t.get("token", "")
symbol = t.get("symbol", "")
name = t.get("name", "")
price_change = t.get("token_price_change_24h", "N/A")
token_list += f"- **{symbol}** ({name}): `{addr}` - 24h change: {price_change}%\n"
response_text = f"Here are the trending tokens on {chain.upper()}:\n\n{token_list}\nWould you like me to set up a strategy for any of these?"
else:
response_text = f"I couldn't find any trending tokens on {chain.upper()}. Try again later."
strategy_update = None
except json.JSONDecodeError:
pass # Use defaults
# Use the native thinking from API if available, otherwise use parsed thinking
final_thinking = thinking or thinking_field
# Check if token_address is missing in strategy_update
strategy_needs_confirmation = False
token_search_results = None
if strategy_update:
# Extract token name from conditions
token_name = None
for cond in strategy_update.get("conditions", []):
if not cond.get("token_address") and cond.get("token"):
token_name = cond.get("token")
strategy_needs_confirmation = True
break
# Search for token if name is found
if strategy_needs_confirmation and token_name:
try:
from ..ave.client import AveCloudClient
from ...core.config import get_settings
settings = get_settings()
ave_client = AveCloudClient(
api_key=settings.AVE_API_KEY,
plan=settings.AVE_API_PLAN
)
# Run async search in sync context
import asyncio
tokens = asyncio.run(ave_client.get_tokens(query=token_name, chain="bsc", limit=5))
if tokens:
token_search_results = [
{
"symbol": t.get("symbol", ""),
"name": t.get("name", ""),
"address": t.get("token", ""), # trending API uses "token" for contract address
"chain": t.get("chain", "bsc")
}
for t in tokens
]
except Exception as e:
print(f"Token search error: {e}")
# Only update strategy if token_address is provided
if strategy_update and strategy_needs_confirmation:
# Don't auto-save - user needs to confirm token address
# Return response but with strategy_update as None
return {
"response": response_text,
"thinking": final_thinking,
"strategy_updated": False,
"strategy_needs_confirmation": True,
"strategy_data": strategy_update,
"token_search_results": token_search_results,
"success": True
}
# Update strategy in database if provided
if strategy_update and self.bot_id:
self._update_strategy(strategy_update)
return {
"response": response_text,
"thinking": final_thinking,
"strategy_updated": strategy_update is not None,
"strategy_needs_confirmation": False,
"success": True
}
except Exception as e:
return {
"response": f"I encountered an error: {str(e)}. Please try again.",
"thinking": None,
"strategy_updated": False,
"success": False
}
def _execute_backtest(
self,
token_address: str,
timeframe: str = "1d",
start_date: str = None,
end_date: str = None
) -> str:
"""Execute a backtest using the bot's current strategy."""
try:
import asyncio
from ...core.database import get_db
from ...db.models import Backtest
from ...services.backtest.engine import BacktestEngine
from ...core.config import get_settings
from datetime import datetime
import uuid
settings = get_settings()
db = next(get_db())
# Get the bot
bot = db.query(Bot).filter(Bot.id == self.bot_id).first()
if not bot:
return "I couldn't find the bot. Please try again."
# Default dates if not provided (last 30 days)
if not end_date:
end_date = datetime.now().strftime("%Y-%m-%d")
if not start_date:
start_date = (datetime.now() - timedelta(days=30)).strftime("%Y-%m-%d")
# Create backtest engine
backtest_config = {
"bot_id": self.bot_id,
"token": token_address,
"chain": "bsc",
"timeframe": timeframe,
"start_date": start_date,
"end_date": end_date,
"strategy_config": bot.strategy_config,
"ave_api_key": settings.AVE_API_KEY,
"ave_api_plan": settings.AVE_API_PLAN,
"initial_balance": 10000.0,
}
engine = BacktestEngine(backtest_config)
results = asyncio.run(engine.run())
# Format results for display
if "error" in results:
return f"Backtest failed: {results['error']}"
total_return = results.get("total_return", 0)
win_rate = results.get("win_rate", 0)
total_trades = results.get("total_trades", 0)
max_drawdown = results.get("max_drawdown", 0)
sharpe_ratio = results.get("sharpe_ratio", 0)
final_balance = results.get("final_balance", 10000)
# Format return with emoji indicators
return_emoji = "📈" if total_return >= 0 else "📉"
return_str = f"+{total_return:.2f}%" if total_return >= 0 else f"{total_return:.2f}%"
drawdown_emoji = "⚠️" if abs(max_drawdown) > 10 else ""
response = f"""Here's the backtest result for {token_address}:
**Performance Summary**
{return_emoji} Total Return: {return_str}
💰 Final Balance: ${final_balance:,.2f}
📊 Total Trades: {total_trades}
🎯 Win Rate: {win_rate:.1f}%
**Risk Metrics**
{drawdown_emoji} Max Drawdown: {max_drawdown:.2f}%
📉 Sharpe Ratio: {sharpe_ratio:.2f}
**Period**: {start_date} to {end_date} ({timeframe})
Would you like me to adjust the strategy parameters based on these results?"""
return response
except Exception as e:
return f"I encountered an error running the backtest: {str(e)}"
def _manage_simulation(
self,
action: str,
token_address: str = None,
kline_interval: str = "1m"
) -> str:
"""Manage trading simulations: start, stop, status, or results."""
try:
import asyncio
import threading
import uuid
from ...core.database import SessionLocal
from ...db.models import Simulation
from ...services.simulate.engine import SimulateEngine
from ...core.config import get_settings
from datetime import datetime
db = SessionLocal()
settings = get_settings()
try:
# Get the bot
bot = db.query(Bot).filter(Bot.id == self.bot_id).first()
if not bot:
return "I couldn't find the bot. Please try again."
if action == "start":
if not token_address:
return "I need a token address to start a simulation. Which token would you like to simulate?"
# Check if there's already a running simulation
running_sim = db.query(Simulation).filter(
Simulation.bot_id == self.bot_id,
Simulation.status == "running"
).first()
if running_sim:
# Stop the existing one first
self._stop_simulation_db(running_sim.id)
# Create new simulation
sim_id = str(uuid.uuid4())
simulation = Simulation(
id=sim_id,
bot_id=self.bot_id,
started_at=datetime.utcnow(),
status="running",
config={
"token": token_address,
"chain": "bsc",
"kline_interval": kline_interval
},
signals=[],
klines=[],
trade_log=[]
)
db.add(simulation)
db.commit()
# Start the simulation in background
sim_config = {
"bot_id": self.bot_id,
"token": token_address,
"chain": "bsc",
"kline_interval": kline_interval,
"max_candles": 100,
"candle_delay": 30 if kline_interval == "1m" else 60,
"strategy_config": bot.strategy_config,
"ave_api_key": settings.AVE_API_KEY,
"ave_api_plan": settings.AVE_API_PLAN,
"initial_balance": 10000.0,
}
# Run simulation in background thread
def run_sim():
asyncio.run(self._run_simulation_sync(sim_id, settings.DATABASE_URL, sim_config))
thread = threading.Thread(target=run_sim)
thread.daemon = True
thread.start()
return f"Started simulation on {token_address} using {kline_interval} klines. The simulation is running and will process up to 100 candles. Ask me for status or results anytime!"
elif action == "stop":
# Find running simulation
running_sim = db.query(Simulation).filter(
Simulation.bot_id == self.bot_id,
Simulation.status == "running"
).first()
if not running_sim:
return "No simulation is currently running."
self._stop_simulation_db(running_sim.id)
# Get final results
portfolio = running_sim.portfolio or {}
current_balance = portfolio.get("current_balance", 10000)
initial_balance = portfolio.get("initial_balance", 10000)
pnl = current_balance - initial_balance
pnl_pct = (pnl / initial_balance) * 100 if initial_balance > 0 else 0
return f"Simulation stopped!\n\nFinal Results:\n💰 Final Balance: ${current_balance:,.2f}\n📈 P&L: {'+' if pnl >= 0 else ''}${pnl:,.2f} ({'+' if pnl_pct >= 0 else ''}{pnl_pct:.2f}%)\n📊 Trades: {len(running_sim.trade_log or [])}"
elif action == "status":
# Find running simulation
running_sim = db.query(Simulation).filter(
Simulation.bot_id == self.bot_id,
Simulation.status == "running"
).first()
if not running_sim:
return "No simulation is currently running."
portfolio = running_sim.portfolio or {}
klines_count = len(running_sim.klines or [])
trade_count = len(running_sim.trade_log or [])
status = f"**Simulation Status: Running**\n\n"
status += f"📊 Candles processed: ~{klines_count}\n"
status += f"📈 Trades executed: {trade_count}\n"
if portfolio.get("position", 0) > 0:
status += f"💰 Position: {portfolio['position']:.4f} {portfolio.get('position_token', 'TOKEN')}\n"
status += f"💰 Cash: ${portfolio.get('current_balance', 0):,.2f}\n"
else:
status += f"💰 Cash: ${portfolio.get('current_balance', 10000):,.2f}\n"
status += "\nAsk me to stop or get full results anytime!"
return status
elif action == "results":
# Find running or most recent simulation
simulation = db.query(Simulation).filter(
Simulation.bot_id == self.bot_id
).order_by(Simulation.started_at.desc()).first()
if not simulation:
return "No simulation found. Start a simulation first!"
portfolio = simulation.portfolio or {}
current_balance = portfolio.get("current_balance", 10000)
initial_balance = portfolio.get("initial_balance", 10000)
pnl = current_balance - initial_balance
pnl_pct = (pnl / initial_balance) * 100 if initial_balance > 0 else 0
trade_log = simulation.trade_log or []
status_emoji = "🟢" if simulation.status == "running" else ""
status_text = "Running" if simulation.status == "running" else "Completed/Stopped"
results = f"**Simulation Results** {status_emoji} ({status_text})\n\n"
results += f"💰 Final Balance: ${current_balance:,.2f}\n"
results += f"📈 P&L: {'+' if pnl >= 0 else ''}${pnl:,.2f} ({'+' if pnl_pct >= 0 else ''}{pnl_pct:.2f}%)\n"
results += f"📊 Total Trades: {len(trade_log)}\n"
if simulation.status == "running":
results += f"\n⏳ Simulation still running... (refresh for latest)"
return results
else:
return f"Unknown action: {action}. Use 'start', 'stop', 'status', or 'results'."
finally:
db.close()
except Exception as e:
return f"I encountered an error managing the simulation: {str(e)}"
def _stop_simulation_db(self, simulation_id: str):
"""Stop a simulation in the database."""
from ...core.database import SessionLocal
db = SessionLocal()
try:
simulation = db.query(Simulation).filter(Simulation.id == simulation_id).first()
if simulation:
simulation.status = "stopped"
db.commit()
finally:
db.close()
async def _run_simulation_sync(self, simulation_id: str, db_url: str, config: dict):
"""Run simulation synchronously in background."""
from ...services.simulate.engine import SimulateEngine
from ...core.database import SessionLocal
async def _run():
engine = SimulateEngine(config)
engine.run_id = simulation_id
def serialize_signal(s):
created = s.get("created_at")
if hasattr(created, "isoformat"):
created = created.isoformat()
return {**s, "created_at": created}
def save_progress():
db = SessionLocal()
try:
sim = db.query(Simulation).filter(Simulation.id == simulation_id).first()
if sim:
sim.status = engine.status
sim.signals = [serialize_signal(s) for s in engine.signals]
sim.klines = [{"time": k.get("time"), "close": k.get("close")} for k in engine.klines]
sim.trade_log = engine.trade_log
sim.portfolio = {
"initial_balance": config.get("initial_balance", 10000),
"current_balance": engine.current_balance,
"position": engine.position,
"position_token": engine.position_token,
"entry_price": engine.entry_price,
"current_price": engine.last_close,
}
db.commit()
finally:
db.close()
try:
await engine.run()
finally:
save_progress()
asyncio.run(_run())
def _update_strategy(self, strategy_update: Dict) -> bool:
"""Update the bot's strategy in the database."""
try:
from ...core.database import get_db
db = next(get_db())
bot = db.query(Bot).filter(Bot.id == self.bot_id).first()
if bot:
bot.strategy_config = strategy_update
db.commit()
return True
except Exception as e:
print(f"Error updating strategy: {e}")
return 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)