Improve timeout and retry handling

[KUNDAN1334]

solved #7

[Copilot]

Pull Request Overview

This PR adds timeout and retry handling capabilities to the langchain-gradient library. The changes implement configurable timeout and retry behavior for API requests, along with proper error handling and documentation.

• Adds custom exception classes (APITimeoutError, APIRetryError) for better error handling
• Implements a RetryConfig class with exponential backoff support
• Updates ChatGradient to support timeout and retry configuration through the Gradient SDK
• Adds comprehensive test coverage for timeout and retry scenarios

Reviewed Changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tests/unit_tests/test_timeout_retry.py	New test file covering timeout configuration, retry logic, and error handling scenarios
langchain_gradient/retry_config.py	New module implementing retry configuration with exponential backoff
langchain_gradient/exceptions.py	New module defining custom exception classes for timeout and retry errors
langchain_gradient/chat_models.py	Updated to support timeout/retry configuration and enhanced error handling with gradient SDK exceptions
docs/timeout_retry-examples.md	New documentation file with usage examples for timeout and retry configuration

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The APIRetryError exception is used but not imported. This will cause a NameError at runtime when the max retries are exceeded.

[Copilot]

The docstring indicates timeout defaults to 60.0 seconds, but the class-level docstring at line 73 states it 'Defaults to 60.0 seconds (Gradient SDK default)'. The actual Gradient SDK default should be verified, as the code now explicitly sets 60.0 instead of allowing the SDK to use its own default. Consider using None if you want to rely on the SDK's default, or clarify that this is now a library-specific default.

Suggested change

	"""Timeout for requests in seconds. Defaults to 60.0 seconds."""
	"""Timeout for requests in seconds. Defaults to 60.0 seconds (library default)."""

[Copilot]

The gradient module is imported twice: once at line 6 as from gradient import Gradient and again at line 7 as import gradient. The second import is only used for exception handling. Consider organizing imports more clearly or adding a comment explaining why both import styles are needed.

[Copilot]

Missing markdown code fence markers around the Python code. The code examples should be wrapped in triple backticks with the python language identifier for proper rendering.

Suggested change

	from langchain_gradient import ChatGradient
	from langchain_gradient.exceptions import APITimeoutError, APIRetryError
	Initialize with timeout and retry configuration
	llm = ChatGradient(
	model="llama3.3-70b-instruct",
	timeout=30.0,
	max_retries=3
	)
	try:
	response = llm.invoke("Hello world!")
	print(response)
	except APITimeoutError as e:
	print(f"Request timed out: {e}")
	except APIRetryError as e:
	print(f"Max retries exceeded: {e}")
	## Advanced Configuration
	from langchain_gradient import ChatGradient
	from langchain_gradient.retry_config import RetryConfig
	Custom retry configuration with exponential backoff
	custom_retry = RetryConfig(
	max_retries=5,
	initial_delay=1.0,
	max_delay=60.0,
	exponential_base=2.0,
	jitter=True
	)
	llm = ChatGradient(
	model="llama3.3-70b-instruct",
	timeout=60.0,
	retry_config=custom_retry
	)
	```python
	from langchain_gradient import ChatGradient
	from langchain_gradient.exceptions import APITimeoutError, APIRetryError
	# Initialize with timeout and retry configuration
	llm = ChatGradient(
	model="llama3.3-70b-instruct",
	timeout=30.0,
	max_retries=3
	)
	try:
	response = llm.invoke("Hello world!")
	print(response)
	except APITimeoutError as e:
	print(f"Request timed out: {e}")
	except APIRetryError as e:
	print(f"Max retries exceeded: {e}")

Advanced Configuration

from langchain_gradient import ChatGradient
from langchain_gradient.retry_config import RetryConfig

# Custom retry configuration with exponential backoff
custom_retry = RetryConfig(
    max_retries=5,
    initial_delay=1.0,
    max_delay=60.0,
    exponential_base=2.0,
    jitter=True
)

llm = ChatGradient(
    model="llama3.3-70b-instruct",
    timeout=60.0,
    retry_config=custom_retry
)

[Copilot]

This return statement is unreachable because the for loop will either return successfully or raise an exception. Remove this line to improve code clarity.

Suggested change

return None