llm-catcher 0.2.2

A Python library that uses LLMs to diagnose and explain exceptions in real-time

LLM Catcher

A Python library that uses LLMs to diagnose and explain exceptions in real-time. It provides intelligent error diagnosis for Python applications, with special support for FastAPI.

Features

• Intelligent exception diagnosis using OpenAI's GPT models
• FastAPI middleware for automatic error handling
• Support for both caught and uncaught exceptions
• Custom error handlers and prompts
• Environment-based configuration
• Comprehensive stack trace analysis
• Schema-aware error diagnosis for FastAPI

Installation

# Install from PyPI
pip install llm-catcher

# For development
pip install llm-catcher[dev]

Quick Start

1. Create a .env file with your OpenAI API key:

LLM_CATCHER_OPENAI_API_KEY=your-api-key-here

2. Choose your exception handling mode:

# Only handle uncaught exceptions (default)
LLM_CATCHER_HANDLED_EXCEPTIONS=UNHANDLED

# Or handle all exceptions
LLM_CATCHER_HANDLED_EXCEPTIONS=ALL

# Or specify exact exceptions
LLM_CATCHER_HANDLED_EXCEPTIONS=ValueError,TypeError,ValidationError

Minimal Example

import asyncio
import traceback
from llm_catcher import LLMExceptionDiagnoser

async def main():
    diagnoser = LLMExceptionDiagnoser()
    try:
        1/0  # Cause an error
    except Exception as e:
        stack_trace = ''.join(traceback.format_exception(type(e), e, e.__traceback__))
        print(await diagnoser.diagnose(stack_trace))

if __name__ == "__main__":
    asyncio.run(main())

FastAPI Integration

from fastapi import FastAPI
from llm_catcher.middleware import LLMCatcherMiddleware
from llm_catcher.settings import Settings

app = FastAPI()

# Configure with Settings object
settings = Settings(
    handled_exceptions=["UNHANDLED"],  # Only handle uncaught exceptions
    ignore_exceptions=["KeyboardInterrupt"],
    custom_handlers={
        "ValueError": "This is a validation error. Please check: \n1. Input types\n2. Required fields",
        "ZeroDivisionError": "This is a division by zero error. Check division operations."
    }
)

# Add the middleware
app.add_middleware(LLMCatcherMiddleware, settings=settings)

# Custom handler for specific exceptions (respected in UNHANDLED mode)
@app.exception_handler(ValueError)
async def value_error_handler(request: Request, exc: ValueError):
    return JSONResponse(
        status_code=400,
        content={"message": "Custom handler: Invalid value provided"}
    )

Standalone Usage

import asyncio
from llm_catcher import LLMExceptionDiagnoser
import traceback

async def main():
    # Initialize diagnoser
    diagnoser = LLMExceptionDiagnoser()

    try:
        # Your code here
        result = 1 / 0
    except Exception as e:
        # Get full stack trace
        stack_trace = ''.join(traceback.format_exception(type(e), e, e.__traceback__))
        # Get AI-powered diagnosis
        diagnosis = await diagnoser.diagnose(stack_trace)
        print(diagnosis)

if __name__ == "__main__":
    asyncio.run(main())

Configuration

All settings can be configured through environment variables or the Settings class:

Required Settings

• LLM_CATCHER_OPENAI_API_KEY: Your OpenAI API key

Optional Settings

• LLM_CATCHER_LLM_MODEL: Model to use (default: "gpt-4")
  • Supported: "gpt-4", "gpt-3.5-turbo", "gpt-4-1106-preview"
• LLM_CATCHER_TEMPERATURE: Model temperature (default: 0.2, range: 0-1)
• LLM_CATCHER_HANDLED_EXCEPTIONS: Which exceptions to handle
• LLM_CATCHER_IGNORE_EXCEPTIONS: Exceptions to ignore
• LLM_CATCHER_CUSTOM_HANDLERS: Custom prompts for specific exceptions

Exception Handling Modes

UNHANDLED Mode (Default)

LLM_CATCHER_HANDLED_EXCEPTIONS=UNHANDLED

• Only handles exceptions that aren't caught by other handlers
• Respects existing exception handlers
• Perfect for adding AI diagnosis without disrupting existing error handling

ALL Mode

LLM_CATCHER_HANDLED_EXCEPTIONS=ALL

• Handles all exceptions (except those in ignore_exceptions)
• Provides AI diagnosis even for caught exceptions
• Useful when you want AI diagnosis for every error

Specific Exceptions

LLM_CATCHER_HANDLED_EXCEPTIONS=ValueError,TypeError,ValidationError

• Only handles listed exception types
• Can be combined with custom handlers

Development

Setup Development Environment

# Clone the repository
git clone https://github.com/yourusername/llm-catcher.git
cd llm-catcher

# Create and activate virtual environment
python -m venv venv
source venv/bin/activate  # Linux/Mac
# or
.\venv\Scripts\activate  # Windows

# Install development dependencies
pip install -e ".[dev]"

Running Tests

# Run all tests
pytest

# Run with coverage
pytest --cov=llm_catcher

# Run specific test file
pytest tests/test_settings.py -v

Examples

The examples/ directory contains two examples demonstrating different use cases:

1. CLI Example (cli_example.py):

   • Demonstrates both UNHANDLED and ALL modes
   • Shows difference between caught and uncaught exceptions
   • Shows standard error handling vs LLM diagnosis
   • Includes stack trace handling

   # Run CLI example
   python examples/cli_example.py
   

2. FastAPI Example (fastapi_example.py):

   • Shows FastAPI middleware integration
   • Demonstrates custom error handlers
   • Includes Pydantic schema validation
   • Shows different error scenarios and their handling
   • Includes custom prompts for specific errors

   # Run FastAPI example
   python examples/fastapi_example.py
   # Then test with curl or browser at http://localhost:8000/docs
   

Each example includes detailed comments and demonstrates best practices for using LLM Catcher in different contexts. The FastAPI example also includes Swagger documentation accessible through the /docs endpoint.

Notes

• API key is required and must be provided via environment or settings
• Settings are validated on initialization
• Invalid values fall back to defaults
• Environment variables take precedence over direct configuration
• Custom handlers must be valid JSON when provided via environment
• Stack traces are included in LLM prompts for better diagnosis

License

This project is licensed under the MIT License - see the LICENSE file for details.