Getting started

Route your first request through Osyra. This guide assumes you have:

• A terminal with curl
• An Osyra account (sign up here)

1. Create a workspace

A workspace is the tenant boundary for API keys and authenticated model requests. Most teams start with one workspace per environment (dev, staging, prod).

Open Workspaces, select New workspace, and provide a name. A handle and description are optional. The console does not ask you to invent or hard-code a deployment region during this flow.

2. Mint an API key

Open API keys and select Create key. Give the key a descriptive name and choose a tier: Sandbox, Test, or Live.

Sandbox is the default self-service tier. Test and Live are long-lived classes and require workspace-admin authorization. The tier defines API-key lifecycle and quota behavior. See Authentication.

The key appears once. Copy it into a secret store (Vault, AWS Secrets Manager, Doppler) — Osyra never displays it again.

3. Make your first call

Osyra exposes a narrow OpenAI-compatible Chat Completions endpoint. The examples below use the supported string-message fields; they do not imply compatibility with every OpenAI request option. See Making requests for the exact mapped subset.

import os
from openai import OpenAI
 
client = OpenAI(
    base_url="https://api.osyra.ai/v1",
    api_key=os.environ["OSYRA_API_KEY"],
)
 
response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {"role": "system", "content": "You are a terse assistant."},
        {"role": "user", "content": "Summarize HTTP/2 in one sentence."},
    ],
)
 
print(response.choices[0].message.content)
print("Response ID:", response.id)

A successful response uses the supported OpenAI Chat Completions response shape and can include additive Osyra metadata. The current compatibility handler exposes usage/model/provider headers; it does not emit an X-Osyra-Receipt header. The response id identifies the model response and must not be treated as a signed receipt.

4. Verify in the dashboard

Open the dashboard Activity panel to inspect recent real events when the Edge activity feed is configured. Event delivery is asynchronous; this guide does not promise a fixed sub-second arrival time.

• The recorded action and outcome
• The event time and correlation identifier
• The emitting service and resource reference when available

Osyra does not promise that raw prompts or responses are present in activity rows. Privacy controls such as no_store can deliberately prevent body persistence.

5. Add a policy

Before production, review the policies actually published for the workspace and test both allow and deny paths with real requests. Policy capabilities and syntax live in Governance & policy; do not assume a control is active because a console page exists.

What's next

• Authentication — key scopes, rotation, revocation, BYOK to providers
• Making requests — chat, embeddings, streaming, retries
• Model routing — the verified chat and embeddings dispatch contract
• Error reference — selected shared Edge codes and recovery guidance