Riven Studio

The prompt and model playground. Draft prompts, compare models side by side, tune parameters, and export the request as working code.

Service

surfaces_studio

URL

studio.rivenai.io

Studio is where a prompt gets built before it becomes code. Iterate interactively, run the same prompt against several models at once, and copy out a request you can paste into your application.

What it gives you#

• Side-by-side comparison — one prompt, several models, responses and latency next to each other.
• Parameter control — temperature, top_p, max_tokens, stop sequences, and reasoning effort, all live.
• System prompt drafting — edit the system message separately from the conversation so it can be versioned.
• Code export — emit the equivalent cURL, Python, or JavaScript call against api.rivenai.io.
• Token and cost preview — see what a request will spend before wiring it into production.

From Studio to production#

Studio runs against the same gateway and the same models as your application, so behaviour carries over. When a prompt is good, export it and move to the API reference; when you need to know whether a change made things better, promote it into a scored suite in Riven Evals.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.rivenai.io/v1",
  apiKey: process.env.RIVEN_API_KEY,
});

const resp = await client.chat.completions.create({
  model: "riven-core",
  messages: [
    { role: "system", content: "You are a precise research assistant." },
    { role: "user", content: "Summarize this quarter's filings." },
  ],
  temperature: 0.4,
});