Answers about SmarToken accounts, model APIs, billing and production use

SmarToken uses Google sign-in for normal user accounts. Existing users sign in with Google, and new Google users can be created automatically on first sign-in when registration is open.

Google sign-in reduces password handling for the public console. API traffic still uses server-side API keys, so user sessions and model calls remain separate.

Sign in, open the console and create a key in the API key area. The full key is shown once. Store it in a server-side secret manager or environment variable, not in browser code.

Yes. Revoke the exposed key, rotate the secret in every service that used it and review recent usage logs for unexpected traffic.

Set the SDK baseURL to https://thesmartoken.com/v1 and authenticate with Authorization: Bearer YOUR_SMARTOKEN_KEY. Chat requests use the standard /chat/completions path under that base URL.

Use a public SmarToken model ID from the model catalog, such as deepseek-chat, kimi-k2 or qwen-plus. Some OpenRouter-style aliases are normalized for migration pages, but production clients should prefer the documented SmarToken ID.

Yes. Send stream: true for server-sent event streaming or stream: false for a normal JSON response. Streaming is useful for chat UIs and developer tools that show partial output.

Yes. Keep the same OpenAI-compatible client and change the model field when you want to compare DeepSeek, Qwen, Kimi, GLM, Doubao or another configured route.

No. It is a machine-readable API response for clients. Human-readable model information lives on /models and individual /models/{id} pages.

Usage is estimated from input and output tokens, priced by model route and deducted from wallet balance after the request or stream completes.

Requests can be rejected when the account lacks enough usable balance. Add balance, reduce output length or choose a lower-cost model route before retrying.

Promotional API credit is usage balance for model calls and is not withdrawable cash unless a specific campaign policy says otherwise.

Use the pricing page for current public model prices and each model detail page for input and output token rates, context notes and source links.

Available payment methods depend on operator configuration. The product can support channels such as Alipay, WeChat Pay, Stripe or crypto deposit flows when enabled.

Check that the Authorization header uses Bearer plus the full SmarToken API key, confirm the key has not been revoked and make sure the request is sent from server-side code.

A 429 usually means rate limiting or temporary load protection. Reduce concurrency, add backoff and include timestamps, model ID and request path if you open a support ticket.

Long prompts, long generated answers, retries and streaming responses can increase token usage. Compare input size, max output settings and actual usage logs for the request.

Routes depend on upstream provider availability and configured keys. Retry with backoff, switch to a comparable model or contact support if the issue affects production traffic.

Include account email, request time, model ID, endpoint path, error code, approximate prompt size and whether the issue happened with streaming or non-streaming mode. Do not include private API keys.

No. Keep keys on your server, worker or secure automation environment. Frontend apps should call your backend, and your backend should call SmarToken.

Run the same evaluation set across candidate models and compare quality, latency, cost and failure modes. Pick a route for a workload, not for the whole company by default.

Review important routes after major provider releases, pricing updates or quality regressions. Keep a small benchmark set from real support, coding or document tasks.

Keep private console, support and API endpoints out of public navigation where possible, block machine endpoints in robots.txt and make sure public docs provide crawlable alternatives.