Calypso RAG is a hosted retrieval workspace for teams that need answers backed by explicit sources. Load knowledge, configure the default agent or named profiles, validate behavior in Playground, then deploy the same grounded agent through Code, the Upload API, n8n, or the Web widget.
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.CALYPSO_PROJECT_API_KEY,
baseURL: "https://YOUR_CALYPSO_BASE_URL/v1"
});
const response = await client.responses.create({
model: "calypso-rag-agent",
input: "Summarize the grounded knowledge available in this workspace."
});
console.log(response.output_text);
The Calypso RAG operating loop
Load knowledge
Curate files, websites, Q&A, and legacy intents inside Knowledge so every grounded answer starts from a clear source set.
Configure the agent
Use Agent Builder to control retrieval scope, role, presentation, small talk, and whether grounded source titles appear in replies.
Validate in Playground
Test the saved agent with a real project API key, inspect grounding states, and tighten behavior before you connect the agent to production surfaces.
Deploy through integrations
Ship the same grounded agent through OpenAI-compatible Code endpoints, the Upload API, n8n, or the website widget without rebuilding policy per surface.
Review outcomes
Use Insights to monitor widget analytics, ratings, quick answers, and no-reply patterns while staying honest about what the current RAG surface measures today.
Start from the real product loop: knowledge, agent policy, Playground validation, integrations, project administration, and the widget analytics that exist today.
