Reason with Tenzro Cortex

The simplest entry point — pick a model, give it a question, and a tier.

import { TenzroClient } from "tenzro-sdk";
const client = new TenzroClient({ endpoint: "https://rpc.tenzro.network" });

const reply = await client.cortex.reason(
  "qwen-3-0.6b",
  "How many TEE-attested validators were active in the last epoch?",
  "standard",
);
console.log(reply);

For production runs, use the full request shape to bound spend, loops, and deadline.

const reply = await client.cortex.reasonWithRequest({
  model_id: "qwen-3-0.6b",
  input: "Summarize the last 5 governance proposals.",
  tier: "standard",
  min_loops: 1,
  max_loops: 5,
  max_cost_tnzo: "0.05",
  deadline_ms: 30_000,
});

Agent memory persists across requests; recall happens with vector kNN, BM25, or hybrid RRF.

const record = await client.memory().grant({
  agent_did: "did:tenzro:machine:...",
  text: "alice prefers Qwen 3 for code review",
  source: "Controller",
});
console.log(record.id);

Cortex workers register on the network; list them locally or via the gossip-learned set.

const local = await client.cortex.listWorkers();
const remote = await client.cortex.listRemoteWorkers();
console.log({ local, remote });