openai_react_agent_service.py

import json
import numpy as np
import pandas as pd
import re
import os
import uuid
import logging
import time
import threading
from io import StringIO
import sys
import traceback
from typing import Optional, Dict, Any, List, Set
from pydantic import BaseModel, Field
from dotenv import load_dotenv
import seaborn as sns 
import datetime as dt
from langchain_openai import ChatOpenAI

# Configure pandas display options
pd.set_option('display.max_columns', None)
pd.set_option('display.max_rows', None)
pd.set_option('display.max_colwidth', None)

# Load environment variables
load_dotenv()

# Configuration constants
API_KEYS = os.getenv("OPENAI_API_KEYS", "").split(",")
MODEL_NAME = 'gpt-4o'
KEY_RETRY_DELAY = 40  # seconds

# Configure non-interactive matplotlib backend
os.environ['MPLBACKEND'] = 'agg'
import matplotlib.pyplot as plt
plt.show = lambda: None  # Disable display

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(levelname)s - %(message)s',
)
logger = logging.getLogger(__name__)

def handle_out_of_range_float(value):
    """Handle NaN and Inf values in numeric data"""
    if isinstance(value, float):
        if np.isnan(value):
            return None
        elif np.isinf(value):
            return "Infinity"
    return value

class OpenAIKeyManager:
    """Manage multiple OpenAI API keys with validation, failover, and delayed retries"""
    
    def __init__(self, api_keys: List[str]):
        self.original_keys = api_keys.copy()
        self.available_keys = api_keys.copy()
        self.active_key = None
        self.failed_keys: Dict[str, float] = {}  # key: timestamp when failed
        self.llm_instance = None
        self.lock = threading.Lock()
        
    def configure(self) -> bool:
        """Validate and activate an OpenAI API key with retry logic"""
        with self.lock:
            # First try available keys
            while self.available_keys:
                key = self.available_keys.pop(0)
                if self._try_key(key):
                    return True
            
            # Then check if any failed keys are ready for retry
            now = time.time()
            retry_keys = [
                k for k, ts in self.failed_keys.items() 
                if (now - ts) >= KEY_RETRY_DELAY
            ]
            
            for key in retry_keys:
                if self._try_key(key):
                    del self.failed_keys[key]
                    return True
            
            logger.critical("All API keys failed (including retries)")
            return False
    
    def _try_key(self, key: str) -> bool:
        """Attempt to use a specific key, return True if successful"""
        try:
            self.llm_instance = ChatOpenAI(
                model=MODEL_NAME,
                api_key=key,
                temperature=0,
                max_retries=0
            )
            self.llm_instance.invoke("test")  # Simple test call
            self.active_key = key
            logger.info(f"Active_Key: {self._mask_key(key)}")
            return True
        except Exception as e:
            self.failed_keys[key] = time.time()
            logger.error(f"Key failed: {self._mask_key(key)} - {str(e)}")
            return False
    
    def rotate_key(self) -> bool:
        """Rotate to the next available API key (including retries)"""
        return self.configure()
    
    def get_llm_instance(self) -> ChatOpenAI:
        """Get the configured LLM instance"""
        return self.llm_instance
    
    def _mask_key(self, key: str) -> str:
        """Mask API key for secure logging"""
        return f"{key[:8]}...{key[-4:]}" if key else ""

class PythonREPL:
    """Secure Python REPL environment for code execution"""
    
    def __init__(self, df: pd.DataFrame):
        self.df = df
        self.local_env = {
            "pd": pd,
            "df": self.df.copy(),
            "plt": plt,
            "os": os,
            "uuid": uuid,
            "sns": sns,
            "json": json,
            "dt": dt,
            "np": np,
        }
        os.makedirs('generated_charts', exist_ok=True)
        
    def execute(self, code: str) -> Dict[str, Any]:
        """Execute Python code in a secure environment"""
        old_stdout = sys.stdout
        sys.stdout = mystdout = StringIO()
        error_msg = None
        
        try:
            # Ensure proper matplotlib configuration
            code = f"""
import matplotlib.pyplot as plt
plt.switch_backend('agg')
{code}
plt.close('all')
            """
            exec(code, self.local_env)
            self.df = self.local_env.get('df', self.df)
            error = False
        except Exception as e:
            error_msg = traceback.format_exc()
            error = True
        finally:
            sys.stdout = old_stdout
            
        return {
            "output": mystdout.getvalue(),
            "error": error,
            "error_message": error_msg if error else None,
            "df": self.local_env.get('df', self.df)
        }

class RethinkAgent(BaseModel):
    """AI agent for data analysis with automatic error correction"""
    
    df: pd.DataFrame
    max_retries: int = Field(default=5, ge=1)
    current_retry: int = Field(default=0, ge=0)
    repl: Optional[PythonREPL] = None
    key_manager: Optional[OpenAIKeyManager] = None
    llm: Optional[ChatOpenAI] = None
    
    class Config:
        arbitrary_types_allowed = True
        
    def _extract_code(self, response: str) -> str:
        """Extract Python code from markdown response"""
        code_match = re.search(r'```python(.*?)```', response, re.DOTALL)
        if code_match:
            return code_match.group(1).strip()
        code_match = re.search(r'```(.*?)```', response, re.DOTALL)
        return code_match.group(1).strip() if code_match else response.strip()
    
    def _generate_initial_prompt(self, query: str, chart: bool = False) -> str:
        """Generate the initial prompt for the LLM"""
        columns = "\n".join([f"{col} ({self.df[col].dtype})" for col in self.df.columns])
        
        if chart:
            return f"""
Generate Python code to create visualization(s) for this DataFrame with columns:
{columns}

First 5 rows:
{self.df.head().to_string()}

Query: {query}

Requirements:
1. Save visualizations to 'generated_charts/' with UUID filename (use uuid.uuid4())
2. Use plt.savefig() with format='png'
3. No plt.show() calls allowed
4. After saving each chart, logger.info exactly: CHART_SAVED: generated_charts/<uuid>.png
5. Start with 'import pandas as pd', 'import matplotlib.pyplot as plt', etc.
6. The DataFrame is available as 'df'
7. Wrap code in ```python``` blocks
8. If Question is illogical and cannot be answered, explain using logger.info()
"""
        else:
            return f"""
Generate Python code to analyze this DataFrame with columns:
{columns}

First 5 rows:
{self.df.head().to_string()}

Query: {query}

Requirements:
1. Use logger.info() to show results with clear explanations
2. If Question is illogical and cannot be answered, explain using logger.info()
3. Start with necessary imports ('import pandas as pd', etc.)
4. The DataFrame is available as 'df'
5. For tabular results, use markdown formatting
6. Wrap code in ```python``` blocks
"""
    
    def _generate_retry_prompt(self, query: str, error: str, code: str, chart: bool = False) -> str:
        """Generate a retry prompt when code execution fails"""
        if chart:
            return f"""
The previous code failed with this error:
{error}

Here was the code that failed:
{code}

Please fix the code to:
1. Create the requested visualization(s)
2. Save to 'generated_charts/' with UUID filename
3. logger.info CHART_SAVED messages
4. Handle the error: {error}

Original query: {query}

Show the corrected code in ```python``` blocks
"""
        else:
            return f"""
The previous code failed with this error:
{error}

Here was the code that failed:
{code}

Please fix the code to:
1. Complete the analysis requested
2. Handle the error: {error}
3. Include clear output formatting

Original query: {query}

Show the corrected code in ```python``` blocks
"""
    
    def initialize_model(self, api_keys: List[str]) -> bool:
        """Initialize OpenAI client with key rotation"""
        self.key_manager = OpenAIKeyManager(api_keys)
        if not self.key_manager.configure():
            raise RuntimeError("All API keys failed")
        self.llm = self.key_manager.get_llm_instance()
        return True
    
    def generate_code(self, query: str, error: Optional[str] = None, 
                     previous_code: Optional[str] = None, chart: bool = False) -> str:
        """Generate Python code to answer the query"""
        prompt = self._generate_retry_prompt(query, error, previous_code, chart) if error else self._generate_initial_prompt(query, chart)
        
        try:
            response = self.llm.invoke(prompt)
            return self._extract_code(response.content)
        except Exception as e:
            logger.error(f"API error: {str(e)}")
            if self.key_manager.rotate_key():
                self.llm = self.key_manager.get_llm_instance()
                return self.generate_code(query, error, previous_code, chart)
            raise
    
    def execute_query(self, query: str, chart: bool = False) -> str:
        """Execute the query with automatic error correction"""
        self.repl = PythonREPL(self.df)
        error = None
        previous_code = None
        
        while self.current_retry < self.max_retries:
            try:
                code = self.generate_code(query, error, previous_code, chart)
                result = self.repl.execute(code)
                
                if result["error"]:
                    self.current_retry += 1
                    error = result["error_message"]
                    previous_code = code
                    logger.warning(f"Retry {self.current_retry}/{self.max_retries}")
                else:
                    self.df = result["df"]
                    return result["output"]
            except Exception as e:
                logger.error(f"Critical error: {str(e)}")
                return f"System error: {str(e)}"
        
        return f"Failed after {self.max_retries} retries. Last error: {error}"

def openai_react_chat(csv_url: str, query: str, chart: bool = False) -> Optional[Dict]:
    """Main function to execute data analysis queries"""
    try:
        # Read and validate input data
        df = pd.read_csv(csv_url)
        if df.empty:
            raise ValueError("Empty DataFrame loaded from CSV")
            
        agent = RethinkAgent(df=df)
        
        if not agent.initialize_model(API_KEYS):
            logger.error("Failed to initialize model")
            return None

        result = agent.execute_query(query, chart)
        
        # Process different response types
        if isinstance(result, pd.DataFrame):
            processed = result.apply(handle_out_of_range_float).to_dict(orient="records")
        elif isinstance(result, pd.Series):
            processed = result.apply(handle_out_of_range_float).to_dict()
        elif isinstance(result, list):
            processed = [handle_out_of_range_float(item) for item in result]
        elif isinstance(result, dict):
            processed = {k: handle_out_of_range_float(v) for k, v in result.items()}
        else:
            processed = {"answer": str(handle_out_of_range_float(result))}
        
        logger.info("Analysis completed successfully")
        
        if chart and isinstance(result, str) and result.startswith("CHART_SAVED:"):
           result = result.strip()  # Remove any leading/trailing spaces or newlines
           match = re.search(r'CHART_SAVED:\s*(\S+)', result)
           if match:
              chart_path = match.group(1)
              logger.info("Chart Path:", chart_path)
              return chart_path
           else:
             logger.info("Could not extract chart path from response")
             return None
        
        return processed
    except Exception as e:
        logger.error(f"Error in openai_llm_chat: {str(e)}")
        return None