feat/Interceptor

[Nisarg38]

Add Request/Response Interceptor Pattern

Overview

This PR introduces a flexible interceptor pattern to the OpenAI Python client, allowing users to intercept and modify requests before they are sent and responses after they are received. This enables powerful use cases like logging, retrying, metrics collection, and custom request/response manipulation.

Key Components

Core Interceptor Framework

• InterceptorRequest: Data class containing request details (method, URL, headers, params, body)
• InterceptorResponse: Generic data class for response data (status code, headers, body, original request)
• Interceptor: Abstract base class defining the interceptor interface
• InterceptorChain: Manages a sequence of interceptors, executing them in order

Example Implementations

Added example interceptors demonstrating practical use cases:

1. LoggingInterceptor: Provides detailed request/response logging
2. RetryInterceptor: Implements exponential backoff retry logic for 5xx errors

Usage Example

client = OpenAI( interceptors=[ LoggingInterceptor(), RetryInterceptor(max_retries=3, initial_delay=1.0) ])

Benefits

• Extensibility: Easy to add new interceptors for custom behavior
• Separation of Concerns: Clean separation between core client logic and auxiliary functionality
• Error Resilience: Built-in error handling in interceptor chain execution
• Type Safety: Full type annotations and generic support

Testing

• Added example usage in examples/interceptor.py
• Tested with logging and retry scenarios
• Verified error handling and chain execution

Future Improvements

• Add more built-in interceptors for common use cases
• Consider adding async support
• Add interceptor ordering/priority mechanism

Breaking Changes

None. This is a new feature that maintains backward compatibility.