@kamiyo-org/eigenai
v0.1.0
Published
EigenAI inference client with KAMIYO escrow protection. Verifiable LLM outputs with quality-based payment settlement.
Maintainers
Readme
@kamiyo-org/eigenai
Escrow-protected AI inference on Solana.
npm install @kamiyo-org/eigenaiUsage
import { Connection, Keypair } from '@solana/web3.js';
import { createKamiyoEigenAI, PROGRAM_IDS } from '@kamiyo-org/eigenai';
const client = createKamiyoEigenAI({
connection: new Connection('https://api.mainnet-beta.solana.com'),
wallet: Keypair.fromSecretKey(/* your key */),
programId: PROGRAM_IDS.MAINNET,
eigenAiAuth: { type: 'apiKey', apiKey: 'your-eigenai-key' },
});
const result = await client.inferenceWithEscrow(
{
model: 'gpt-oss-120b-f16',
messages: [{ role: 'user', content: 'Hello' }],
escrowAmount: 0.01,
},
userTokenAccount,
treasury
);
if (result.success) {
console.log(result.response);
}Flow
- Lock SOL in escrow
- EigenAI returns response + attestation
- Release funds or dispute
Auth
// API Key
eigenAiAuth: { type: 'apiKey', apiKey: 'key' }
// Wallet Grant
eigenAiAuth: { type: 'grant', privateKey: Uint8Array(32), walletAddress: '0x...' }API
// Escrow-protected inference
client.inferenceWithEscrow(params, userTokenAccount, treasury)
// Release with rating (1-5)
client.releaseEscrow(escrowId, 5)
// Dispute
client.disputeWithAttestation(escrowId)
// Get dispute evidence
client.getDisputeEvidence(escrowId) // { attestation, prompt, output }
// Direct call (no escrow)
client.callEigenAI(params)Quality Tiers
| Score | Refund | |-------|--------| | 80+ | 0% | | 65-79 | 35% | | 50-64 | 75% | | <50 | 100% |
Config
createKamiyoEigenAI({
connection,
wallet,
programId,
eigenAiAuth,
defaultEscrowAmount: 0.01, // SOL
defaultQualityThreshold: 70,
defaultTimeoutMs: 60000,
debug: false,
})Constants
PROGRAM_IDS.MAINNET // FVnvAs8bahMwAvjcLq5ZrXksuu5Qeu2MRkbjwB9mua3u
PROGRAM_IDS.DEVNET // EqScj2SUahLLUuP56s77yK6bPr3VEPoTyDecjvyoBtxT
KAMIYO_MINT // Gy55EJmheLyDXiZ7k7CW2FhunD1UgjQxQibuBn3Npump
LIMITS.MIN_ESCROW_SOL // 0.001
LIMITS.MAX_ESCROW_SOL // 100
LIMITS.MAX_MESSAGES // 100
LIMITS.SESSION_ID_LENGTH // 32Models
gpt-oss-120b-f16qwen3-32b-128k-bf16
License
MIT
