Tase API
One endpoint for chat, tool calling, and streaming. The wire format is OpenAI-compatible, so most existing clients work by changing the base URL and key.
Getting started
- OverviewThe Tase API gives you programmatic access to the Tase model through an OpenAI-compatible chat completions endpoint.
- QuickstartGet an API key, make your first chat completion in curl, JavaScript, or Python, and switch to streaming.
- AuthenticationHow Tase API keys work: the tase_sk_ format, the Authorization header, handling 401 errors, and rotating keys safely.
Models
- ModelsThe Tase model family: what each model is built for, how to choose one, and how to list available models through the API.
- tase-8bModel card for tase-8b: a compact model trained by Tase Industries, specialized for tool calling with low latency in Turkish and English.
- Pricing and limitsHow quotas and rate limits work on the Tase API: requests per minute, concurrency, 429 handling with Retry-After, fair use, and usage tracking.
Guides
- Chat completionsSend messages to Tase models and read the response: roles, system prompts, parameters, multi-turn conversations, and token usage.
- Streaming responsesStream chat completions over server-sent events: enable stream mode, parse the SSE wire format, accumulate deltas, handle [DONE], and cancel mid-flight.
- Structured outputsGet reliable JSON from Tase models: JSON mode with response_format, schema instructions in the system message, validation with retries, and when to use tool calling instead.
- Function callingConnect the Tase model to your own functions: define tools, receive structured tool calls, execute them in your app, and feed results back for a final answer.
- Built-in toolsThe tool families trained into the Tase model — tasks, health, finance, memory, agents, connections, and world knowledge — and how to use them without sending a single schema.
- Tool use patternsProduction patterns for tool calling with the Tase API: feeding results back, handling errors and empty results, confirming destructive actions, and running multi-step workflows.
API reference
- Create chat completionFull reference for POST /v1/chat/completions: request body fields, response schema, streaming format, and tool calling examples.
POST /v1/chat/completions - List modelsReference for GET /v1/models: list the model IDs available to your API key and read each model object.
GET /v1/models - ErrorsHow Tase API errors are shaped, what every HTTP status code means, and how to retry safely with exponential backoff.
- Rate limitsHow Tase API rate limits work: the 429 response, the Retry-After header, backoff, concurrency, and batching strategies.
Cookbook
- SDKs and librariesCall the Tase API from JavaScript, Python, Swift, or plain curl. No dedicated SDK required — any OpenAI-compatible client works with a base URL change.
- Build an assistant with toolsAn end-to-end tool-calling loop in Node.js: take a user message, let the model call your functions, feed results back, and return the final answer.
- Structured data extractionTurn free-form text into validated JSON: describe the schema in the system message, parse defensively, and retry once with the error fed back.
- Migrating to TaseMove an existing OpenAI-compatible integration to the Tase API: change the base URL, the key, and the model name — everything else stays.