Integrate quiz and RAG

1. 1. Create an Edy account and generate a key at Dashboard - API keys.
2. 2. Ask an Edy administrator to provision the key for QUIZ and RAG routes.
3. 3. Store the key in your backend secret manager as EDY_API_KEY.

Never put the key in browser JavaScript, a mobile app, source control, or a public frontend environment variable.

Use the deployment URL as the base URL and send the key as a Bearer token. The alternative X-API-Key header is also accepted.

const baseUrl = "https://edycode.vercel.app";
const headers = {
  Authorization: `Bearer ${process.env.EDY_API_KEY}`,
  "Content-Type": "application/json",
};

Send extracted document text. Edy does not upload or extract files for this endpoint.

const response = await fetch(`https://edycode.vercel.app/api/v1/quiz/generate`, {
  method: "POST",
  headers,
  body: JSON.stringify({
    document_text: documentText,
    question_count: 5,
    question_types: ["multiple_choice", "true_false", "short_answer"],
    difficulty: "medium",
  }),
});

if (!response.ok) throw new Error(JSON.stringify(await response.json()));
const quiz = await response.json();

Supported counts are 5 and 10. Supported types are multiple_choice, true_false, and short_answer. The returned questions include validated answers and explanations.

Your application owns document storage and retrieval. Build this pipeline in your backend:

1. Extract text from the source document.
2. Split it into chunks and assign stable IDs.
3. Generate embeddings and store them in your vector database.
4. Retrieve the best chunks for each query.
5. Send those chunks to Edy as trusted context data.

const response = await fetch(`https://edycode.vercel.app/api/v1/rag/answer`, {
  method: "POST",
  headers,
  body: JSON.stringify({
    query: "What causes osmosis?",
    context_chunks: [
      {
        chunk_id: "biology-001-chunk-03",
        text: "Water moves from higher water potential to lower water potential during osmosis.",
        retrieval_score: 0.91,
      },
    ],
    conversation_history: [],
    stream: false,
  }),
});

const answer = await response.json();
// answer.citations contains only submitted chunk IDs.

Edy answers only from the supplied chunks. When the context is insufficient, the response has refused: true, an empty citations array, and low confidence.

Set stream to true and parse Server-Sent Events on your server. Render answer.delta events as they arrive, but treat the completion event as the authoritative final result.

event: answer.delta
data: {"delta":"Osmosis occurs...","request_id":"..."}

event: completion
data: {"answer":"Osmosis occurs...","citations":["biology-001-chunk-03"],"confidence":"high","refused":false}

• Keep all Edy requests behind your backend.
• Log the Edy X-Request-Id with your own request ID.
• Do not persist full document text or retrieved chunks in application logs.
• Use stable chunk IDs so citations can link back to your source document.
• Enforce your own tenant authorization before sending context to Edy.