Skip to main content
If you serve many end-users, tag each call with your user id. Orbitrage then breaks down cost, usage, and flow graphs per customer — inside one project. It’s your own data about your own users.

​
One user per process

import orbitrage
orbitrage.init(api_key, user_id=current_user.id)

​
A multi-tenant server

Set the user at the start of each request, then build the client. 
import os, orbitrage
orbitrage.init(os.environ["ORBITRAGE_API_KEY"])

from fastapi import FastAPI, Request
from openai import OpenAI
app = FastAPI()

@app.post("/chat")
def chat(request: Request, body: dict):
    orbitrage.set_user(body["user_id"])     # tag this request's calls
    client = OpenAI()                        # construct AFTER set_user
    resp = client.chat.completions.create(
        model="grok-4-fast",
        messages=body["messages"],
    )
    return resp.choices[0].message.content
In Python, construct the OpenAI() client after set_user() — already-built clients keep their original headers. Node’s withAssociationProperties scopes the user to the block, so any client built inside it is tagged.

​
What you get

In the dashboard, filter any workflow by end-user to see:
  • spend and call volume per customer,
  • their per-run flow graph,
  • and churn / frustration signals in the Intelligence layer.
Set only x-orbitrage-end-user-id (what user_id does). Never send x-orbitrage-user-id — that’s the gateway’s account identity and returns 403.