Developer Mode & Advanced Debugging

Developer Mode is an advanced set of workflows and utilities designed to help engineers debug API interactions, inspect raw payloads, and safely test applications without burning production credits.

1. Using Test Keys

The foundation of Developer Mode is the cbx_test_ API key.

When you create a key in the dashboard and set its environment to Test, ChatboxAI changes its internal routing behavior:

• Zero Cost: Requests made with cbx_test_ keys do not deduct from your USD credit balance (if you are on a Paid plan) and do not count against your 300 requests/day limit (if you are on a Free plan).
• Mock Routing: Depending on the requested model, ChatboxAI may route the request to a smaller, free-tier equivalent (e.g., routing a chatboxai/llama-3.3-70b request to chatboxai/llama-3.1-8b) to fulfill the request for free.
• Strict Throttling: Test keys are hard-capped at 5 Requests Per Minute (RPM) to prevent abuse.

Use cbx_test_ keys in your local .env files and CI/CD pipelines.

2. Inspecting Raw Requests

Sometimes standard SDKs abstract away too much information. If you need to debug exactly what headers and body payloads are reaching ChatboxAI, you can enable verbose logging in your SDK, or use cURL with the -v flag.

Debugging with cURL

curl -v -X POST https://chatboxai.co.in/api/v1/chat/completions \
  -H "Authorization: Bearer cbx_test_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "chatboxai/llama-3.1-8b", "messages": [{"role": "user", "content": "Test"}]}'

Watch the output for the X-RateLimit-* and X-API-Credits-Remaining response headers to verify your authentication context.

3. Simulating Errors

To ensure your application handles edge cases gracefully, you can force the ChatboxAI API to return specific errors by injecting special strings into the model parameter using a cbx_test_ key.

(Note: This feature is currently in Beta)

Example

{
  "model": "dev-mock-429",
  "messages": [{"role": "user", "content": "Hello"}]
}

If sent with a cbx_test_ key, this request will instantly return a 429 Too Many Requests error, allowing you to test if your retry loop works properly before pushing to production.