Quickstart — First Inference Call

This guide gets you from zero to a working streaming inference call as fast as possible.

1. Create an Account & Get an API Key

1. Go to https://hypervize.tech and sign in (or create an account) using Auth0.

   • A default inference-scoped API key is automatically generated for you on signup (via database trigger).

2. If needed, go to Settings → Keys (or directly to /dashboard/settings/keys) to view it or create additional keys.

3. Give any new key a name (e.g., “Production – Elastic”) and choose the inference scope.

4. Copy the key immediately. It will look like:

   TEXT

   COPY

   hvz_live_3f8a9c2e1b4d5f6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1b2c

Important: Treat this key like a password. It grants access to inference on your behalf.

You now have everything needed for the Elastic Inference API.

2. Make Your First Call (cURL)

Replace YOUR_API_KEY with the key you just generated.

curl https://hypervize.tech/api/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4.6",
    "messages": [
      {"role": "user", "content": "Explain scale-to-zero inference in one paragraph."}
    ],
    "max_tokens": 256,
    "stream": true
  }'

You should see Server-Sent Events (SSE) chunks starting with data: {...} and ending with data: [DONE].

3. Try a Different Model

The model field accepts the display name from our catalog (recommended) or the raw provider identifier.

Popular starting models:

• claude-sonnet-4.6
• llama-3-3-70b-instruct
• mistral-large-3
• deepseek-v3.2
• qwen3-32b

See the full list in Supported Models.

4. (Optional) Use the Dashboard Playground

1. Go to Dashboard → Inference.
2. Switch to the Elastic tab.
3. Select a model from the grouped dropdown.
4. Type a prompt and hit Send.

The playground shows token usage, TTFB, and total latency — very useful while you’re learning the catalog.

5. Next: Try a Dedicated Endpoint

Once you’re comfortable with Elastic:

1. Go to Dashboard → Inference.
2. Switch to the Dedicated tab.
3. Enter a Hugging Face model ID (e.g., meta-llama/Llama-3.1-8B-Instruct).
4. Configure min/max instances and deploy.

After provisioning completes (status changes to ONLINE), you can call it at:

https://hypervize.tech/api/d/{endpoint-id}/chat/completions

Dedicated endpoints support the identical request format as Elastic.

Troubleshooting First Calls

What’s Next?

• Read the full Authentication & API Keys guide
• Deep dive into Elastic Inference API
• Learn how to deploy Dedicated Inference Endpoints