Early access: features, availability, and pricing may change as Kyara Intelligence evolves.

Reference

Unsupported features

Kyara Intelligence is text in, text out. These are the OpenAI features, endpoints and parameters it doesn't offer.

Kyara supports chat completions for text generation. Any OpenAI-compatible client can connect, but the capabilities below are not available.

Summary

  • Function and tool calling: tools and tool_choice are dropped, so the model never sees them.
  • Structured outputs and JSON mode: response_format is dropped.
  • Vision: image inputs are rejected. Message content must be plain text.
  • Audio: no audio input or output, speech synthesis or transcription.
  • Embeddings: there is no /v1/embeddings endpoint.

Available endpoints

These are the endpoints for API clients. Every other path returns 404 in the OpenAI error envelope.

EndpointStatus
POST /v1/chat/completionsAvailable (text generation only)
GET /v1/modelsAvailable (no key needed)
GET /v1/balanceAvailable (Kyara extension)
/v1/completions (legacy)Not available
/v1/embeddingsNot available
/v1/images/*Not available
/v1/audio/*Not available
/v1/responsesNot available
/v1/files, /v1/batchesNot available
/v1/fine_tuning/*Not available
/v1/moderationsNot available
/v1/assistants, /v1/threads, /v1/vector_storesNot available
/v1/realtimeNot available

Request parameters

Chat completions requests are checked against an allowlist. Every parameter is either passed to the model, dropped without an error, or rejected with 400.

Supported

Accepted and passed to the model:

  • model, messages
  • temperature, top_p, top_k, min_p, presence_penalty, frequency_penalty, repetition_penalty, seed
  • max_tokens, max_completion_tokens, stop (up to four sequences)
  • n (only 1)
  • stream, stream_options (include_usage, include_obfuscation)
  • logprobs, top_logprobs, user, session_id
  • reasoning (effort, max_tokens, enabled, exclude)
  • reasoning_effort (folded into reasoning.effort; an explicit reasoning.effort wins)

Dropped without an error

  • tools, tool_choice, functions, parallel_tool_calls
  • response_format (JSON mode and structured outputs)
  • logit_bias, prediction
  • modalities, audio, web_search_options
  • store, metadata, service_tier

Rejected with an error

These return a 400 validation error in the OpenAI envelope, with param naming the field; see Validation errors.

  • Message content that isn't a plain string, such as a content array with image_url or audio parts.
  • A message role other than system, developer, user or assistant. There is no tool role.
  • Unknown fields inside a message, stream_options or reasoning. Messages accept only role, content, name, reasoning, reasoning_content and reasoning_details, so an assistant message carrying tool_calls is rejected.
  • An n other than 1, more than four stop sequences, or a session_id longer than 256 characters.

Notes

  • top_k, min_p, repetition_penalty and session_id are Kyara extensions beyond the OpenAI standard; see Chat completions.
  • Requests without max_tokens or max_completion_tokens get an output limit of 8,192 tokens, and higher limits than the model allows are lowered to its maximum.
  • Support for sampling parameters varies by model.