0G Compute Network guide for decentralized AI inference, fine-tuning, and GPU services. Covers chatbots, image generation, speech-to-text, SDK integration (0g-serving-broker), processResponse API, broker.inference methods, CLI commands (0g-compute-cli), and account management. Use this skill for any 0G compute, 0G AI, or decentralized GPU question.
npx skills add https://github.com/internet-court/internet-court-skill --skill 0g-compute
This skill provides instructions for building with the 0G Compute Network — a decentralized GPU marketplace for AI inference and model fine-tuning. Follow these patterns exactly when generating code.
processResponse() after every API response (see processResponse section below).When unsure about a pattern, reference the detailed guides:
| Network | RPC URL | Inference | Fine-tuning |
|---------|---------|-----------|-------------|
| Mainnet | https://evmrpc.0g.ai | Yes | Yes |
| Testnet | https://evmrpc-testnet.0g.ai | Yes | Yes |
Model availability changes frequently. Always use broker.inference.listService() or 0g-compute-cli inference list-providers to check current models. On-chain model names use org/model-name format.
node --version # Must be >= 22.0.0
pnpm add @0glabs/0g-serving-broker # SDK for applications
pnpm add @0glabs/0g-serving-broker -g # CLI for direct usage
0g-compute-cli setup-network # Choose testnet or mainnet
0g-compute-cli login # Login with wallet private key
0g-compute-cli deposit --amount 10 # Deposit funds
0g-compute-cli get-account # Check balance
import { ethers } from "ethers";
import { createZGComputeNetworkBroker } from "@0glabs/0g-serving-broker";
const RPC_URL = process.env.NODE_ENV === 'production'
? "https://evmrpc.0g.ai"
: "https://evmrpc-testnet.0g.ai";
const provider = new ethers.JsonRpcProvider(RPC_URL);
const wallet = new ethers.Wallet(process.env.PRIVATE_KEY!, provider);
const broker = await createZGComputeNetworkBroker(wallet);
// Discover services
const services = await broker.inference.listService();
services.forEach(s => {
console.log(`${s.provider} | ${s.model} | ${s.serviceType}`);
});
// Make inference request
const { endpoint, model } = await broker.inference.getServiceMetadata(providerAddress);
const headers = await broker.inference.getRequestHeaders(providerAddress);
const response = await fetch(`${endpoint}/chat/completions`, {
method: "POST",
headers: { "Content-Type": "application/json", ...headers },
body: JSON.stringify({ messages, model })
});
const data = await response.json();
// Extract chatID (see chatID table below)
let chatID = response.headers.get("ZG-Res-Key") || response.headers.get("zg-res-key");
if (!chatID) chatID = data.id;
// CRITICAL: Always call processResponse
await broker.inference.processResponse(
providerAddress, // 1st: provider address
chatID, // 2nd: response identifier for verification
JSON.stringify(data.usage) // 3rd: usage data for fee calculation
);
For streaming, browser SDK, cURL, and Python examples, see references/inference.md.
Call broker.inference.processResponse() after EVERY API response for fee settlement and TEE verification.
await broker.inference.processResponse(
providerAddress, // 1st: provider address
chatID, // 2nd: response identifier for verification
JSON.stringify(data.usage) // 3rd: usage data for fee calculation
);
Parameter order: provider, chatID, usageData. Do NOT reorder.
Always try ZG-Res-Key response header first. Use fallback only when header is absent.
| Service Type | chatID Source | Fallback |
|---|---|---|
| Chatbot | ZG-Res-Key header | data.id from response body |
| Text-to-Image | ZG-Res-Key header | none |
| Speech-to-Text | ZG-Res-Key header | none |
| Chatbot Streaming | ZG-Res-Key header | id from stream chunk |
| Audio Streaming | ZG-Res-Key header | none |
Fine-tuning is available on both mainnet and testnet. It is a 6-step CLI process: list providers, upload dataset, calculate tokens, create task, monitor, download and decrypt.
For the complete workflow, see references/fine-tuning.md.
The 0G Compute Network uses Main Accounts (deposits/withdrawals) and Provider Sub-Accounts (service payments). Sub-account refunds have a 24-hour lock period.
0g-compute-cli get-account # Check balance
0g-compute-cli deposit --amount 10 # Deposit to main
0g-compute-cli transfer-fund --provider <ADDR> --amount 5 # Transfer to sub-account
0g-compute-cli retrieve-fund # Retrieve from sub (24h lock)
0g-compute-cli refund --amount 5 # Withdraw to wallet
For detailed account management, see references/account-management.md.
# Inference
0g-compute-cli inference list-providers # List all providers
0g-compute-cli inference verify --provider <ADDR> # Verify TEE attestation
0g-compute-cli inference acknowledge-provider --provider <ADDR> # Required before first use
0g-compute-cli inference get-secret --provider <ADDR> # Get API key for direct calls
0g-compute-cli inference serve --provider <ADDR> --port 3000 # Local OpenAI-compatible proxy
# Fine-tuning
0g-compute-cli fine-tuning list-providers # List fine-tuning providers
0g-compute-cli fine-tuning list-models # List available models
# Web UI
0g-compute-cli ui start-web # Launch at localhost:3090
| Problem | Solution |
|---|---|
| Insufficient balance | deposit --amount 5 then transfer-fund --provider <ADDR> --amount 2 |
| Provider not acknowledged | inference acknowledge-provider --provider <ADDR> |
| Provider busy (fine-tuning) | Wait and retry, or choose a different provider |
| Web UI port conflict | ui start-web --port 3091 |
> Note: A unified skill covering all 0G services (Compute, Storage, Chain) exists at 0g-agent-skills.
Convert PyTorch AT_DISPATCH macros to AT_DISPATCH_V2 format in ATen C++ code. Use when porting AT_DISPATCH_ALL_TYPES_AND*, AT_DISPATCH_FLOATING_TYPES*, or other dispatch macros to the new v2 API. For ATen kernel files, CUDA kernels, and native operator implementations.
Write docstrings for PyTorch functions and methods following PyTorch conventions. Use when writing or updating docstrings in PyTorch code.
Statistical models library for Python. Use when you need specific model classes (OLS, GLM, mixed models, ARIMA) with detailed diagnostics, residuals, and inference. Best for econometrics, time series, rigorous inference with coefficient tables. For guided statistical test selection with APA reporting use statistical-analysis.
Answer questions about the AI SDK and help build AI-powered features. Use when developers: (1) Ask about AI SDK functions like generateText, streamText, ToolLoopAgent, embed, or tools, (2) Want to build AI agents, chatbots, RAG systems, or text generation features, (3) Have questions about AI providers (OpenAI, Anthropic, Google, etc.), streaming, tool calling, structured output, or embeddings, (4) Use React hooks like useChat or useCompletion. Triggers on: "AI SDK", "Vercel AI SDK", "generateText", "streamText", "add AI to my app", "build an agent", "tool calling", "structured output", "useChat".
Create an llms.txt file from scratch based on repository structure following the llms.txt specification at https://llmstxt.org/
Use when working directly with the `esm` Python SDK, ESM3 or ESMC model IDs, Forge/Biohub inference clients, or ESMFold2 folding workflows.
Modal is a serverless cloud platform for running Python on demand, including on-demand GPUs. Use when deploying or serving AI/ML models, running GPU-accelerated workloads (training, fine-tuning, inference), serving web endpoints, scheduling batch jobs, or scaling Python code to cloud containers with the Modal SDK.
Use Therapeutics Data Commons through the PyTDC Python package for registry discovery, approved dataset access, task-aware splits, evaluator metrics, benchmark groups, and bounded molecular-oracle workflows.
Take internet-court/0g-compute from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.
The instructions reference pnpm.
Without those the skill loads but fails at the first command.