@foundryprotocol/0gkit-compute

@foundryprotocol/0gkit-compute

Neutral 0G Compute: provider discovery, broker inference, and an OpenAI-compatible shim. Built on @foundryprotocol/0gkit-core. The @0gfoundation/0g-compute-ts-sdk and ethers are optional peers.

Install

npm install @foundryprotocol/0gkit-compute @foundryprotocol/0gkit-core viem
npm install @0gfoundation/0g-compute-ts-sdk ethers@6.13.1
# ethers is pinned to 6.13.1 by @0gfoundation/0g-storage-ts-sdk; using the same
# version here keeps the two SDKs installable side by side.

Use

import { Compute } from "@foundryprotocol/0gkit-compute";

const compute = new Compute({ brokerKey, provider });
const { output, receipt } = await compute.inference({
  messages: [{ role: "user", content: "Hello 0G" }],
});

// Or drop-in OpenAI-style:
const oa = compute.openai();
const res = await oa.chat.completions.create({
  messages: [{ role: "user", content: "Hello 0G" }],
});

Estimating & dry-run

const est = await compute.estimate({
  messages: [{ role: "user", content: "What is 2+2?" }],
  maxOutputTokens: 64,
});
// { kind: "compute", gas: 0n, fee, breakdown: { inputTokens, outputTokensMax, model } }

const dr = await compute.inference(
  { messages: [{ role: "user", content: "ping" }] },
  { dryRun: true }
);
// { dryRun: true, estimate, result: { output: "", receipt: { latencyMs: 0 }, raw } }

Token counts use a chars / 4 heuristic (OpenAI's documented English approximation). Estimates are order-of-magnitude — precise tokenizers add megabytes of vocab files for sub-cent precision. Use 0g estimate compute --prompt "..." --max-output 64 from the CLI.

License

MIT.