Batch inference

Run many prompts as encrypted inferences against the same worker pool - results in submission order, a stalled slot fails on its own. Run it live with your wallet below (user-pays, one job per prompt), or copy the parallel Node call.

NetworkMainnetchain 9200

Build the call

Prompts

Concurrency (Node code)4 in flightMax inferences in flight in the generated call (1-8). The live demo runs sequentially - your wallet signs one at a time.Shared system prompt (optional)

Results

Run the batch to see each prompt's decrypted answer, in order.

Each slot is one createSession + submitJob pair on Mainnet (real LCAI).

The same call in Node (parallel)

TypeScript

import { runInferenceBatch } from "lightnode-sdk";

const results = await runInferenceBatch({
  network: "mainnet",
  privateKey: process.env.PRIVATE_KEY,
  concurrency: 4,
  prompts: [
    "What is Porsche?",
    "What is Lamborghini?",
  ],
  onSlotComplete: ({ index, result, error }) =>
    console.log(index, error?.message ?? result?.answer),
});

// Results are in submission order; a stalled slot fails on its own.
for (const r of results) console.log(r.index, r.error?.message ?? r.result?.answer);

Scaffold it into your project: npx lightnode add batch.