OpenAI-compatible. One line to switch.

Authentication

Create a key in Account → API keys. Send it as a bearer token on every request. Keys start with zy_live_. Each key can have its own monthly credit limit; revoke it at any time.

Authorization: Bearer zy_live_…

Never expose keys in client-side code. Proxy requests through your backend.

Quickstart

Base URL: https://zyverna.com/api/v1

curl https://zyverna.com/api/v1/chat/completions \
  -H "Authorization: Bearer $ZYVERNA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "zyverna-core",
    "messages": [{"role": "user", "content": "Explain tail-call optimization in 3 sentences."}]
  }'

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.ZYVERNA_API_KEY,
  baseURL: "https://zyverna.com/api/v1",
});

const res = await client.chat.completions.create({
  model: "zyverna-core",
  messages: [{ role: "user", content: "Write a Zod schema for a User." }],
});

console.log(res.choices[0].message.content);
console.log("credits:", res.zyverna.credits);

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["ZYVERNA_API_KEY"],
    base_url="https://zyverna.com/api/v1",
)

stream = client.chat.completions.create(
    model="zyverna-flash",
    stream=True,
    messages=[{"role": "user", "content": "Summarize this PR description…"}],
)

for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

POST /chat/completions

Accepts the standard OpenAI request body. Supported parameters: model, messages, stream, temperature, top_p, max_tokens / max_completion_tokens, tools, tool_choice, response_format, stop, user. Responses are standard chat.completion objects with an extra zyverna field.

Streaming

Set stream: true to receive Server-Sent Events. The final chunk before [DONE] carries usage and zyverna.credits.

Tool calling

All models support parallel-safe function calling with JSON Schema parameters.

const res = await client.chat.completions.create({
  model: "zyverna-core",
  messages: [{ role: "user", content: "What's the weather in Toronto?" }],
  tools: [{
    type: "function",
    function: {
      name: "get_weather",
      description: "Get current weather for a city",
      parameters: {
        type: "object",
        properties: { city: { type: "string" } },
        required: ["city"],
      },
    },
  }],
});

const call = res.choices[0].message.tool_calls?.[0];
// → { function: { name: "get_weather", arguments: '{"city":"Toronto"}' } }

Models & pricing

GET /models returns the list below with live pricing. Prices are in credits per 1,000 tokens; 100 credits = 1 USD.

Credits & billing

• API usage draws from the same credit balance as the workspace. Included plan credits are used first, then purchased credits.
• Cached prompt tokens are billed at 50% of the input rate.
• Requests that fail with a 4xx error are not charged. Requests cancelled mid-stream are charged for tokens generated.
• When your balance reaches zero, requests return 402 insufficient_credits. Top up in USD, CAD, EUR or GBP.

Rate limits

Limits are returned in X-RateLimit-Limit-Requests, X-RateLimit-Remaining-Requests and Retry-After headers. Exceeding them returns 429.

Errors

Errors use the OpenAI envelope so SDK error handling works unchanged.

{
  "error": {
    "message": "Invalid API key format.",
    "type": "invalid_request_error",
    "code": "invalid_api_key",
    "param": null
  }
}