Public specOpenAI 1.0 compatible

The wire protocol is OpenAI. The settlement is Solana.

A curated tour of the pieces builders integrate against. The full reference lives in wattz-inference-gateway.

Quickstart

Point any OpenAI client at the Wattz gateway. Response envelope and SSE contract are identical to OpenAI 1.0.

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.WATTZ_API_KEY,
  baseURL: 'https://api.wattz.fi/v1',
});

const stream = await client.chat.completions.create({
  model: 'llama-3-8b-instruct',
  messages: [{ role: 'user', content: 'Explain Solana Token-2022 hooks.' }],
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? '');
}

Model registry

Registry entries live as PDAs seeded with model_id. Each entry stores license, checksum, and version. Publishing a model reserves the PDA and pays a small $WATTZ registration fee.

Copy anchor

#[account]
pub struct ModelRegistry {
    pub authority: Pubkey,
    pub model_id: String,       // 'llama-3-8b-instruct'
    pub license: LicenseClass,  // Apache | Llama | MIT | OpenRAIL | Commercial
    pub weights_checksum: [u8; 32],
    pub version: u16,
    pub published_at: i64,
    pub kyc_gated: bool,
}

Compute attestation

Every inference response carries an attestation quote. Clients that need extra assurance may request a Risc0 or SP1 receipt for the tokenizer + sampler path.

Copy anchor

POST /v1/chat/completions
{
  "model": "mistral-7b-instruct-v0.3",
  "messages": [...],
  "wattz": {
    "attestation": "sgx",
    "zk_receipt": "risc0"
  }
}

// response tail
data: {"type":"attestation","attestation":{
  "proofHash":"9ab5...",
  "attestationType":"sgx",
  "verifier":"wattz-compute-verifier v0.9.4",
  "timestamp":1735689600000
}}

Streaming settlement

Callers pre-authorize a $WATTZ escrow. Each token triggers a Token-2022 transfer hook, splitting revenue between node, host, and treasury. Anchor bulk-settles hourly.

Copy anchor

SPLIT
  node:     68%   // GPU operator
  host:     22%   // Model publisher
  treasury: 10%   // Buyback + burn, grants, audits

Operator CLI

npm i -g wattz-cli. The CLI wraps node runtime, key management, and Anchor calls.

Copy anchor

wattz node init --region us-east
wattz node stake 25000
wattz node start --models llama-3-8b,mistral-7b
wattz infer --model llama-3-8b --prompt 'Hello'
wattz payouts --last 7d