Using Your Inference

Using Your Inference

Each creator gets a dedicated inference wallet. Your harvested USDG routes to that wallet, and inference spend is charged against your own wallet balance through an OpenAI-compatible endpoint. Nothing is pooled across creators.

Check your balance

GET https://api.berryfinance.site/api/creators/{address}/balance
-> { "credited6dp": "...", "spent6dp": "...", "remaining6dp": "..." }

Values are USDG with 6 decimals and track your dedicated inference wallet balance. Credited history: GET /api/creators/{address}/inference.

Authenticate with your wallet

There are no API keys. You prove control of your address by signing a message:

berry-inference:{your_address_lowercase}:{timestamp_ms}

Signatures are valid for 5 minutes. Send three headers with each request:

x-berry-address: 0xYourAddress
x-berry-timestamp: 1234567890123
x-berry-sig: 0xSignature

Spend

curl -X POST https://api.berryfinance.site/api/inference/chat \
  -H "x-berry-address: 0xYourAddress" \
  -H "x-berry-timestamp: TS" \
  -H "x-berry-sig: 0xSig" \
  -H "Content-Type: application/json" \
  -d '{"model":"venice-llama70b","messages":[{"role":"user","content":"hello"}]}'

The body is standard OpenAI chat-completions JSON. The response includes a billing object with the estimated cost and your remaining balance. HTTP 402 means no entitlement remains.

Costs are charged as conservative estimates and reconciled against provider-reported costs; over-charges are credited back during reconciliation.