mirror of
https://github.com/crewAIInc/crewAI.git
synced 2026-01-22 14:48:13 +00:00
Implement comprehensive OAuth2 improvements based on code review feedback
- Add custom OAuth2 error classes (OAuth2Error, OAuth2ConfigurationError, OAuth2AuthenticationError, OAuth2ValidationError) - Implement URL and scope validation in OAuth2Config using pydantic field_validator - Add retry mechanism with exponential backoff (3 attempts, 1s/2s/4s delays) to OAuth2TokenManager - Implement thread safety using threading.Lock for concurrent token access - Add provider validation in LLM integration with _validate_oauth2_provider method - Enhance testing with validation, retry, thread safety, and error class tests - Improve documentation with comprehensive error handling examples and feature descriptions - Maintain backward compatibility while significantly improving robustness and security Co-Authored-By: João <joao@crewai.com>
This commit is contained in:
@@ -2,6 +2,74 @@
|
||||
|
||||
CrewAI supports OAuth2 authentication for custom LiteLLM providers through configuration files.
|
||||
|
||||
## Features
|
||||
|
||||
- **Automatic Token Management**: Handles OAuth2 token acquisition and refresh automatically
|
||||
- **Multiple Provider Support**: Configure multiple OAuth2 providers in a single configuration file
|
||||
- **Secure Credential Storage**: Keep OAuth2 credentials separate from your code
|
||||
- **Seamless Integration**: Works with existing LiteLLM provider configurations
|
||||
- **Error Handling**: Comprehensive error handling for authentication failures
|
||||
- **Retry Mechanism**: Automatic retry with exponential backoff for token acquisition
|
||||
- **Thread Safety**: Concurrent access protection for token caching
|
||||
- **Configuration Validation**: Automatic validation of URLs and scope formats
|
||||
|
||||
## Error Handling
|
||||
|
||||
The OAuth2 implementation provides specific error classes for different failure scenarios:
|
||||
|
||||
### OAuth2ConfigurationError
|
||||
Raised when there are issues with the OAuth2 configuration:
|
||||
- Invalid configuration file format
|
||||
- Missing required fields
|
||||
- Unknown OAuth2 providers
|
||||
|
||||
### OAuth2AuthenticationError
|
||||
Raised when OAuth2 authentication fails:
|
||||
- Network errors during token acquisition
|
||||
- Invalid credentials
|
||||
- Token endpoint errors
|
||||
|
||||
### OAuth2ValidationError
|
||||
Raised when configuration validation fails:
|
||||
- Invalid URL formats
|
||||
- Malformed scope values
|
||||
|
||||
### Example Error Handling
|
||||
|
||||
```python
|
||||
from crewai import LLM
|
||||
from crewai.llms.oauth2_errors import OAuth2ConfigurationError, OAuth2AuthenticationError
|
||||
|
||||
try:
|
||||
llm = LLM(
|
||||
model="my_custom_provider/my-model",
|
||||
oauth2_config_path="./litellm_config.json"
|
||||
)
|
||||
result = llm.call("test message")
|
||||
except OAuth2ConfigurationError as e:
|
||||
print(f"Configuration error: {e}")
|
||||
except OAuth2AuthenticationError as e:
|
||||
print(f"Authentication failed: {e}")
|
||||
if e.original_error:
|
||||
print(f"Original error: {e.original_error}")
|
||||
```
|
||||
|
||||
## Configuration Validation
|
||||
|
||||
The OAuth2 configuration is automatically validated:
|
||||
|
||||
- **URL Validation**: `token_url` must be a valid HTTP/HTTPS URL
|
||||
- **Scope Validation**: `scope` cannot contain empty values when split by spaces
|
||||
- **Required Fields**: `client_id`, `client_secret`, `token_url`, and `provider_name` are required
|
||||
|
||||
## Retry Mechanism
|
||||
|
||||
Token acquisition includes automatic retry with exponential backoff:
|
||||
- 3 retry attempts by default
|
||||
- Exponential backoff: 1s, 2s, 4s delays
|
||||
- Detailed logging of retry attempts
|
||||
- Thread-safe token caching
|
||||
|
||||
## Configuration
|
||||
|
||||
Create a `litellm_config.json` file in your project directory:
|
||||
@@ -98,11 +166,12 @@ The `litellm_config.json` file should follow this schema:
|
||||
}
|
||||
```
|
||||
|
||||
## Error Handling
|
||||
## Legacy Error Handling
|
||||
|
||||
- If OAuth2 authentication fails, a `RuntimeError` will be raised with details
|
||||
- Invalid configuration files will raise a `ValueError` with specifics
|
||||
- Network errors during token acquisition are wrapped in `RuntimeError`
|
||||
For backward compatibility, the following error types are still supported:
|
||||
- OAuth2 authentication failures raise `OAuth2AuthenticationError` (previously `RuntimeError`)
|
||||
- Invalid configuration files raise `OAuth2ConfigurationError` (previously `ValueError`)
|
||||
- Network errors during token acquisition are wrapped in `OAuth2AuthenticationError`
|
||||
|
||||
## Examples
|
||||
|
||||
|
||||
Reference in New Issue
Block a user