Development Mode Features
Dev mode enables special features designed for local development and testing:
📝 Verbose Request Logging
Log full request headers, bodies, and response details for debugging API interactions and understanding data flow.
🎭 Mock Response Mode
Return pre-configured mock responses without calling real APIs, useful for offline development and predictable testing.
⚡ Unlimited Rate Limits
Disable or significantly increase rate limits during development to avoid interruptions while iterating on code.
🐛 Error Injection
Simulate API errors, timeouts, and failures to test error handling code paths and resilience patterns.
Configuration
Enabling Dev Mode
# Environment variable
GATEWAY_MODE=development
# Or in configuration file
mode: development
dev_features:
verbose_logging: true
mock_responses: false
bypass_rate_limits: trueComplete Dev Configuration
| Setting | Description | Default |
|---|---|---|
| verbose_logging | Enable detailed request/response logging | true |
| mock_responses | Use mock responses instead of real API calls | false |
| bypass_rate_limits | Disable rate limiting checks | true |
| error_injection_rate | Percentage of requests to inject errors (0-100) | 0 |
| cache_bypass | Skip response caching | true |
| pretty_print_json | Format JSON responses for readability | true |
Mock Response System
Configuring Mock Responses
mock_responses:
enabled: true
fallback_to_real: false
responses:
- endpoint: /v1/chat/completions
method: POST
response:
id: "mock-chat-001"
object: "chat.completion"
choices:
- message:
role: "assistant"
content: "This is a mock response"
finish_reason: "stop"
- endpoint: /v1/models
method: GET
response:
object: "list"
data:
- id: "gpt-4-mock"
object: "model"
owned_by: "mock-system"Dynamic Mock Responses
# Python mock handler
def mock_chat_completion(request):
return {
"id": f"mock-{request['model']}",
"choices": [{
"message": {
"role": "assistant",
"content": f"Mock response for: {request['messages'][-1]['content']}"
}
}]
}Debug Tools
Request Tracing
Enable request tracing to see the complete lifecycle of each API call:
tracing:
enabled: true
include_headers: true
include_body: true
max_body_size: 10000 # CharactersExample Debug Output
2025-03-16 10:23:45 [DEBUG] Incoming request:
Method: POST
Path: /v1/chat/completions
Headers:
Authorization: Bearer sk-test-...
Content-Type: application/json
Body: {"model":"gpt-4","messages":[...]}
2025-03-16 10:23:45 [DEBUG] Routing to provider: openai
2025-03-16 10:23:45 [DEBUG] Provider request:
URL: https://api.openai.com/v1/chat/completions
Headers: {...}
2025-03-16 10:23:47 [DEBUG] Provider response:
Status: 200 OK
Latency: 1.82s
Body: {"id":"chatcmpl-...","choices":[...]}Error Injection Testing
error_injection:
enabled: true
rate: 10 # Inject errors in 10% of requests
errors:
- type: rate_limit_exceeded
probability: 0.4
- type: timeout
probability: 0.3
- type: server_error
probability: 0.3
Never Enable in Production
Dev mode features like rate limit bypass, error injection, and mock responses should never be enabled in production environments. Always use environment-specific configurations.
Performance Profiling
Enable Performance Metrics
profiling:
enabled: true
track_memory: true
track_cpu: true
slow_request_threshold: 1000 # ms
# Export metrics
export:
format: prometheus
endpoint: /metricsPerformance Dashboard
Access the built-in performance dashboard at http://localhost:9090/dashboard when dev mode is enabled to view:
- Request latency distribution
- Error rates by endpoint
- Memory usage over time
- Active connections
- Cache hit/miss rates