Quickstart

This guide takes a new tenant to one successful response from a hosted Platform model. OneNexus operates the model and catalog endpoint, so this path needs no GPU quota.

Before you start

You need:

access to the OneNexus Console with an approved tenant. If you do not have one, request pilot access;
Python 3 with pip available; and
permission to generate an API key.

Make your first call

Choose a Platform model.

In the Console, open Platform models and expand a model row. Copy these two values:
- the shared Endpoint displayed above the catalog; and
- the exact model value inside that row’s cURL request.
Use the generated cURL value rather than the Public model ID column. The latter is the catalog and billing identity; the serving router dispatches on the model string shown in the cURL request.

If the row shows an amber notice that the policy layer has not acknowledged the listing, wait for it to clear or choose another model — a newly published listing can take about a minute to become callable.

The endpoint above the table, and the model value inside the expanded row’s cURL — those two, not the Public model ID column.
Generate a Platform API key.

Open API Keys → Generate API key. Choose an expiry and set Type to Platform.

Copy the secret from the success dialog; the Console shows it only once.

A Platform key shows no model list — it calls the whole catalog and cannot be narrowed.

Set the connection values.

Replace <model-value-from-curl> with the model value copied in step 1. The configured URL below is the normal catalog URL; if the Console shows a different endpoint, use the Console value.

Bash or zsh

export ONX_BASE_URL="https://catalog.onenexus-do.cloud/v1"
export ONX_MODEL="<model-value-from-curl>"

Run the key prompt separately, then paste the key when prompted, to keep the secret out of shell history.

printf "OneNexus Platform API key: "; IFS= read -rs ONX_API_KEY </dev/tty && export ONX_API_KEY && printf "\n"

PowerShell

$env:ONX_BASE_URL = "https://catalog.onenexus-do.cloud/v1"
$env:ONX_MODEL = "<model-value-from-curl>"
$secret = Read-Host "OneNexus Platform API key" -AsSecureString
$env:ONX_API_KEY = [System.Net.NetworkCredential]::new("", $secret).Password
Remove-Variable secret

Install the SDK and create the script.

Install the OpenAI Python SDK:

python -m pip install openai

Save this as first_call.py:

import os

from openai import OpenAI

client = OpenAI(
    base_url=os.environ["ONX_BASE_URL"],
    api_key=os.environ["ONX_API_KEY"],
)

response = client.chat.completions.create(
    model=os.environ["ONX_MODEL"],
    messages=[{"role": "user", "content": "Say hello in one sentence."}],
)

print("Success:", response.choices[0].message.content)

Run the request and confirm the response.
Terminal
```
python first_call.py
```
Success prints a model-generated message prefixed with Success:. For example:
```
Success: Hello! I hope you're having a great day.
```
The wording after Success: varies; a complete final-answer result has response.choices[0].message.content. A 2xx response with empty final text still proves the API was reached; inspect the full response before retrying, since some models can return reasoning without final content.

If the first call fails

Result	What to check
`401`	The `Authorization` header received a key. Generate a replacement if the secret was lost.
`403`	The key Type is Platform, the catalog endpoint is correct, the request uses the exact `model` value from the row’s cURL request, and no policy-layer notice is showing. Do not retry unchanged.
`404`	The URL ends in `/v1/chat/completions`. The catalog does not currently expose `GET /v1/models`.
`429` or `503`	Retry only a bounded number of times with exponential backoff and jitter. Escalate if the error persists.

Do not log the API key or put it in source code. For the complete request contract and error guidance, see the API reference.

Next steps

Previous
Overview Next
API reference