Use prompt templates

Your project admin can publish reusable prompt templates — a named prompt with placeholders. Instead of hard-coding prompt text in your app, you invoke a template by name and pass values for its variables. The gateway fills in the template and sends the full request to the model.

Invoke a template

Call the normal chat-completions endpoint, but send template (the name) and properties (its variables) instead of a messages array:

curl https://<your-gateway>/v1/chat/completions \
  -H "Authorization: Bearer $OPSTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "template": "code-review",
        "properties": { "language": "Go", "diff": "<your diff here>" }
      }'

The gateway substitutes each from properties, attaches the template's model, and forwards a normal request. The response is a standard chat-completion — nothing else in your client changes.

Notes

• Templates are per-project and must be published. Ask your admin which template names exist (they manage them under Projects → Templates). Invoking a name that isn't published for your project returns 404 unknown_template.
• Unfilled variables stay literal. If you omit a , the text is sent as-is — pass every variable the template expects.
• You can still call models normally. Sending a regular messages array (no template field) works exactly as before; templates are opt-in per request.

Next steps

• Models & routing — the logical models you can call.
• Connect a client — point your OpenAI-compatible tool at the gateway.