Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.prudra.dev/llms.txt

Use this file to discover all available pages before exploring further.

​
Handle multi-step workflows

A multi-step session workflow follows this pattern: the agent pays once (first request), receives a session ID, then makes as many subsequent requests as needed — all sharing the same vault and accumulating results.

​
The two-request pattern

Step 1 — First request (pays, creates session): 
curl -i -X POST https://your-api.com/agent/step \
  -H "Content-Type: application/json" \
  -H "Authorization: Payment eyJ..." \
  -d '{"step": "search", "query": "quantum computing papers 2025"}'
Response: 
HTTP/1.1 200 OK
X-PRUDRA-SESSION-ID: sess_clx1abc123
Content-Type: application/json

{
  "vaultId":   "vlt_clx1def456",
  "sessionId": "sess_clx1abc123",
  "step":      "search",
  "result":    { "papers": ["paper1", "paper2", "paper3"] }
}
The agent saves vaultId and sessionId for subsequent requests. Step 2 — Subsequent request (no payment): 
curl -X POST https://your-api.com/agent/step \
  -H "Content-Type: application/json" \
  -H "X-PRUDRA-SESSION-ID: sess_clx1abc123" \
  -d '{"step": "summarise", "query": "quantum computing papers 2025"}'
Response: 
{
  "vaultId":   "vlt_clx1def456",
  "sessionId": "sess_clx1abc123",
  "step":      "summarise",
  "result":    { "summary": "Found 3 papers on quantum computing..." }
}
The vaultId is identical to Step 1 — the same vault. Both steps’ documents are in it.

​
Confirming vault accumulation

After all steps complete, verify that the vault contains all accumulated results: 
curl https://api.prudra.dev/vaults/vlt_clx1def456 \
  -H "Authorization: Bearer prv_test_sk_..."
Response: 
{
  "id":     "vlt_clx1def456",
  "status": "active",
  "documents": [
    {
      "id":      "doc_1",
      "name":    "Step: search",
      "addedAt": "2026-04-30T09:00:01.000Z"
    },
    {
      "id":      "doc_2",
      "name":    "Step: summarise",
      "addedAt": "2026-04-30T09:00:05.000Z"
    }
  ],
  "events": [
    { "event": "agent.step", "payload": { "step": "search" } },
    { "event": "agent.step", "payload": { "step": "summarise" } }
  ]
}
Both documents from both steps are present in the single vault.

​
Full TypeScript example

// From example-04-session-payments.ts

async function runMultiStepSession(serverUrl: string) {
  // Step 1: First request — pays with MPP
  const r1 = await fetch(`${serverUrl}/agent/step`, {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      'X-PAYMENT': 'stub_payment_accepted',  // stub mode for testing
    },
    body: JSON.stringify({ step: 'search', query: 'latest AI papers' }),
  });

  const step1 = await r1.json();
  const sessionId = step1.sessionId;
  const vaultId   = step1.vaultId;

  console.log('Session created:', sessionId);
  console.log('Vault created:', vaultId);

  // Step 2: Reuses session — no payment
  const r2 = await fetch(`${serverUrl}/agent/step`, {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      'X-PRUDRA-SESSION-ID': sessionId,
    },
    body: JSON.stringify({ step: 'summarise', query: 'latest AI papers' }),
  });

  const step2 = await r2.json();
  console.log('Same vault:', step2.vaultId === vaultId); // true

  // Step 3: Seal the session vault
  const r3 = await fetch(`${serverUrl}/agent/finish`, {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      'X-PRUDRA-SESSION-ID': sessionId,
    },
    body: JSON.stringify({}),
  });

  const finish = await r3.json();
  console.log('Session complete:', finish.status);
  console.log('All documents:', finish.documents.length); // 2
}

​
What the vault contains after all steps

// After all steps, read the vault manifest
const manifest = await vault.getManifest();

console.log('Documents:', manifest.documents.length);  // one per step
console.log('Events:', manifest.events.length);         // one per vault.emit() call
console.log('Files:', manifest.files.length);           // any files uploaded during steps

// Each document from each step
for (const doc of manifest.documents) {
  console.log(doc.name, doc.data);
}