On this page

Chat completions

POST https://aispotmarket.com/buy/api/v1/chat/completions

Spot accepts the common OpenAI chat-completions envelope and adds one routing field, max_price. The Gateway removes max_price and forwards the remaining request body to the selected seller. Market admission currently requires string message content, so multipart content arrays and other newer OpenAI shapes are not part of the guaranteed contract.

The live gateway matches the bid against the model book, dispatches to the winning seller, and returns a standard chat-completion body.

Auth: Bearer buyer key.

See also rate limits, errors, and response consumption.

Spot parameter

field	type	notes
`max_price`	decimal string, optional	Maximum price per 1,000 tokens. Requests that cannot fill at or below it return `402`. Omitting it submits an uncapped immediate bid; it does not use the midpoint or an account-level default.
`model`	string, required	An instrument string, not a vendor model name. See Instruments.

region and venue_pin are not chat-routing fields. Region remains a filter for capacity orders and market-data views; seller selection for chat uses the instrument book and max_price.

Request

field	type	notes
`model`	string	Instrument / model id
`messages`	`{role, content}[]`	OpenAI chat turns
`max_price`	decimal string	Max price per 1,000 tokens
`max_tokens`	number	Optional generation cap
`temperature`	number	Optional sampling
`stream`	bool	SSE when `true`

Response

For stream=false, Spot returns the selected seller's OpenAI-shaped chat.completion JSON. For stream=true, Spot streams the seller's SSE bytes without routing them through Market. Spot does not add a top-level response object; use trade history for execution records.

Errors

Call now (account key): no_fill_at_limit, rate_limited, seller_unreachable, insufficient_balance, unauthorized.

Spend (ticket credential): contract_exhausted, seller_default, contract_expired, contract_not_yet_deliverable. Status codes on Errors.

cURL

curl -s http://127.0.0.1:8080/v1/chat/completions \
  -H "Authorization: Bearer sk-buyer" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4","messages":[{"role":"user","content":"hello"}],"stream":false,"max_price":"0.05","temperature":0.2,"max_tokens":128}'

Rust

use token_gateway_sdk::{BuyerChatMessage, BuyerClient, BuyerKey, ChatCompletionsRequest};
// {"model":"gpt-4","messages":[{"role":"user","content":"hello"}],"stream":false,"max_price":"0.05","temperature":0.2,"max_tokens":128}

let client = BuyerClient::new("http://127.0.0.1:8080", BuyerKey("sk-buyer".into()));
let response = client
    .chat_completions(&ChatCompletionsRequest {
        model: "gpt-4".into(),
        messages: vec![BuyerChatMessage {
            role: "user".into(),
            content: "hello".into(),
        }],
        stream: false,
        max_price: None,
        temperature: None,
        max_tokens: Some(128),
    })
    .await?;
let _ = response.choices[0].message.content.clone();

Python

# {"model":"gpt-4","messages":[{"role":"user","content":"hello"}],"stream":false,"max_price":"0.05","temperature":0.2,"max_tokens":128}
from token_gateway import BuyerAuth, BuyerTokenGateway

gw = BuyerTokenGateway("http://127.0.0.1:8080", buyer=BuyerAuth("sk-buyer"))
print(gw.chat_completions(
    model="gpt-4",
    messages=[{"role": "user", "content": "hello"}],
    max_price="0.05",
    temperature=0.2,
    max_tokens=128,
))

TypeScript

// {"model":"gpt-4","messages":[{"role":"user","content":"hello"}],"stream":false,"max_price":"0.05","temperature":0.2,"max_tokens":128}
import { BuyerClient } from "@ai-token-gateway/sdk";

const buyer = new BuyerClient({ baseUrl: "http://127.0.0.1:8080", apiKey: "sk-buyer" });
const res = await buyer.chatCompletions({
  model: "gpt-4",
  messages: [{ role: "user", content: "hello" }],
  max_price: "0.05",
  temperature: 0.2,
  max_tokens: 128,
});
console.log(res.choices[0].message.content);