Skip to main content

Request Types

The Resultity API supports multiple inference request types to accommodate a wide range of use cases — from fast single-turn queries to heavy distributed tasks and multimodal jobs.

Currently Available​

Synchronous​

  • Standard request-response cycle.
  • The client sends a request and waits for the result in the same HTTP connection.
  • Ideal for fast, low-latency jobs (e.g., short completions).

Polling​

  • The client sends a job and immediately receives a request_id.
  • The result must be fetched later using a status endpoint.
  • Designed for longer or queued tasks.

Planned​

Batch​

  • Submit multiple requests in a single call.
  • Suitable for preloading embeddings, generating multiple completions, etc.

Streaming​

  • Sends tokens to the client as they are generated.
  • Used for chatbots, streaming UIs, and voice assistants.

Multimodel​

  • Executes the same prompt on multiple models in parallel.
  • Can be used for ensemble approaches or fallback logic.
  • Examples: mistral + gemma, or openchat + zephyr.

Multinode​

  • Executes a single heavy model (e.g., LLaMA2 70B) across several nodes using sharding.
  • Enables distributed inference with large context windows.

Media-Oriented Extensions​

Planned support for non-text jobs using compatible APIs:

Vision​

  • /v1/images/generations — generate images from prompts (DALL·E-style);
  • /v1/images/variations — modify or enhance existing images;
  • /v1/images/description — describe or caption images;
  • Based on models such as Kandinsky, Stable Diffusion, Playground v2.

Audio​

  • /v1/audio/transcriptions — convert speech to text (e.g., Whisper, SeamlessM4T);
  • /v1/audio/speech — text-to-speech (e.g., Bark, xtts, tts-zero);
  • /v1/audio/translation — audio-based translation.

Video (experimental)​

  • Future support for generation and description (e.g., SVD, AnimateDiff, Zer0Scope);
  • Initially via Space or custom Subclouds, not main Cloud API.

This modular structure allows Resultity to evolve from OpenAI-compatible endpoints to full multimodal capability, leveraging existing open-source and hosted models while maintaining API consistency.