Reference

API reference

Everything the gateway serves, in one place.

Base URL & CORS

Prefix	Auth	CORS	Purpose
https://api.thairouter.ai/v1	API key	any origin, no credentials	OpenAI-compatible surface
https://api.thairouter.ai/public	none	any origin	Read-only, aggregate catalog data
https://api.thairouter.ai/health	none	—	Liveness

The dashboard, auth and Stripe routes (/dashboard, /api/auth, /admin, /stripe) are session-authenticated for the web app and are not part of the public API.

Chat completions

POST/v1/chat/completionsBearer sk-tr-…

Create a completion, optionally streamed. Full guide: Chat completions.

Field	Type	Description
modelrequired	string	Live ThaiRouter model id.
messagesrequired	array	Non-empty chat history.
max_tokens	integer	Omitted = full output budget. Clamped to the model cap and context room.
reasoning_effort / reasoning	string / object	Thinking level and `exclude`. See Reasoning.
stream	boolean	SSE with a final usage chunk.Default `false`.
…	any	Other OpenAI fields forwarded as-is. See Parameters.

Responds 200 with a chat completion (plus thairouter.usage_id, thairouter.cost and thairouter.reasoning_effort), or text/event-stream with x-thairouter-usage-id and x-thairouter-reasoning-effort headers. Errors: Errors.

List models

GET/v1/modelsBearer sk-tr-…

Live models only, OpenAI list shape with context_length, max_output_tokens, pricing, reasoning (levels, default, whether it can be switched off; null if none) and supported_parameters added.

{
  "object": "list",
  "data": [
    { "id": "thairouter/glm-5.3-flash", "object": "model", "created": 0, "owned_by": "thairouter",
      "context_length": 393216, "max_output_tokens": 131072,
      "pricing": { "promptPerMillion": 5, "completionPerMillion": 15 },
      "reasoning": { "efforts": ["low", "high", "max"], "default": "max", "optional": false },
      "supported_parameters": ["max_tokens", "temperature", "top_p", "stop", "seed", "frequency_penalty",
        "presence_penalty", "response_format", "stream", "reasoning", "reasoning_effort"] }
  ]
}

Public catalog

GET/public/modelsno auth

Every model, including coming-soon ones, with the full catalog entry (description, tags, benchmarks), all-time aggregate stats and a 7-day performance snapshot. Cached for 60 s.

{
  "models": [
    {
      "id": "thairouter/glm-5.3-flash",
      "name": "GLM 5.3 Flash",
      "provider": { "id": "zai", "name": "Z.ai", "url": "https://z.ai" },
      "status": "live",
      "contextLength": 393216,
      "maxOutputTokens": 131072,
      "pricing": { "promptPerMillion": 5, "completionPerMillion": 15 },
      "tags": ["open-weights", "reasoning", "long-context"],
      "stats": { "requests": 1234, "tokens": 5678901, "lastUsedAt": "2026-09-10T08:00:00.000Z" },
      "perf": {
        "windowDays": 7, "requests": 800, "errors": 3, "successRate": 0.996,
        "latencyP50Ms": 1450, "latencyP95Ms": 6200, "tokensPerSec": 62.4
      }
    }
  ]
}

perf is null when a model had no traffic in the window.

Public activity

GET/public/activity?model=thairouter/glm-5.3-flash&days=30no auth

Field	Type	Description
modelrequired	query string	Any catalog id. Unknown ids return 404.
days	query integer	1 to 90.Default `30`.

Daily aggregates across all users. Days with no traffic are filled with zeros. Cached for 5 minutes.

{
  "model": "thairouter/glm-5.3-flash",
  "days": 30,
  "series": [
    { "day": "2026-08-12", "requests": 41, "prompt": 30210, "completion": 12880, "latencyMs": 1390 },
    { "day": "2026-08-13", "requests": 0,  "prompt": 0,     "completion": 0,     "latencyMs": null }
  ]
}

Health

GET/healthno auth

{ "ok": true }

GET / returns { "name": "thairouter-api", "ok": true }. Any other unknown path returns 404 with type: "not_found".