# LLM Request Flow

This is the normal engine user flow.

## Quickstart flow

1. start the server  
2. keep your API key where it already is  
3. point your IDE at `http://localhost:8080`  
4. use your IDE normally  
5. check health with `cast doctor` or the runtime endpoints

## Request path

1. your IDE sends the request to Cencurity Engine  
2. the engine forwards the request to the configured upstream provider  
3. the provider starts returning model output  
4. the engine inspects the response in the stream  
5. the engine returns the final result to your IDE

## Runtime endpoints

- Health: `http://localhost:8080/healthz`  
- Metrics: `http://localhost:8080/metrics`

## Auth behavior

In the normal setup:

- your IDE or coding client already has the provider API key  
- Cencurity Engine forwards that auth header upstream  
- you do not need to store the key in Cencurity Engine