Skip to main content

Quickstart

Connect a local /v1/chat/completions endpoint to an external testing platform through spectral-bridge.

Start a local endpoint​

tip

If you already have a service exposing POST /v1/chat/completions, you can follow along with that instead. Just note its local endpoint (e.g., http://192.168.1.100:8000).

Install FastAPI:

Example
pip install "fastapi[standard]"

Save this echo bot as echo_bot.py. It mirrors the last user message back as the assistant reply, which is enough to validate the connection before switching to your real system.

Example
from fastapi import FastAPI, Request
from fastapi.responses import JSONResponse


app = FastAPI()




@app.post("/v1/chat/completions")
async def chat(messages: list[dict]):
    last = messages[-1]["content"]
    return JSONResponse(
        content={
            "choices": [
                {"message": {"role": "assistant", "content": last}}
            ]
        }
    )

Start it:

Example
fastapi run echo_bot.py --port 8000

The server listens on http://localhost:8000.

Connect to the testing platform​

Install spectral-bridge

Example
pip install "spectral-bridge[pass-through]"

Set your API key

The testing platform provides an API key that authenticates the relay client with their relay server. Set it as an environment variable:

Example
export SPECTRAL_BRIDGE_API_KEY=<your-api-key>

Connect to the relay

Replace --relay-url with the URL from your testing platform. If your endpoint runs on a different host or port, update --target accordingly.

Example
spectral-bridge start \
  --relay-url  <relay-server-url> \
  --adapter    pass-through \
  --target     <local-target-url>

Replace values in angle brackets with your specific configuration:

  • <relay-server-url>: the WebSocket URL of the relay server provided by your testing platform (e.g., wss://bridge.spectral.principled.app).
  • <local-target-url>: the URL of your local endpoint (e.g., http://localhost:8000, http://192.168.1.100:8000).

Once the connection is established, leave it running as long as you want to allow the testing platform to send requests to your AI system through spectral-bridge. Once you terminate it, the connection is closed and the testing platform can no longer reach your system.

Next steps​