Specifics of API implementation in Yandex Cloud AI Studio
Yandex Cloud AI Studio offers a broad choice of APIs you can use to build various applications and address your business needs with the help of AI models. The APIs fall into two groups:
- OpenAI-compatible APIs providing a unified interface for models, text and voice agents, files, and search.
- Specialized APIs developed by Yandex Cloud for text and image generation, classification, model fine-tuning, and batch processing.
OpenAI-compatible APIs
Choose the OpenAI-compatible APIs to create agents, implement RAG scenarios, and for simple model prompting.
| API | Technology | Description |
|---|---|---|
| Models API | REST | Getting a list of available models and their IDs |
| Chat Completions API | REST | Prompt-based text generation without dialog management |
| Conversations API | REST | Working with dialog history and context |
| Responses API | REST | Creating agents. Allows you to generate text, call tools, get structured responses, implement RAG scenarios, and create multi-agent systems |
| Realtime API | WebSocket | Low-latency streaming generation of text and audio for voice-input scenarios |
| Files API | REST | File upload, storage and management for RAG scenarios |
| Embeddings API | REST | Obtaining embeddings for semantic tasks |
| Vector Store API | REST | Data indexing and extraction for RAG |
Specialized Yandex Cloud APIs
Use specialized AI Studio APIs for the YandexGPT Lite, YandexGPT Pro, and YandexART models, batch processing, text classification, model fine-tuning, and MCP server management.
Yandex Cloud APIs are designed based on gRPC. The API methods and data structures are described using Protocol Buffers (proto 3).
Unlike the APIs of most Yandex Cloud services, model APIs do not support the resource-based approach because they do not operate resources. Queries to AI Studio models are not idempotent.
| API | Technology | Description |
|---|---|---|
| Text Generation API | gRPCREST | Prompt-based text generation, function invocation, structured response |
| Image Generation API | gRPCREST | Image generation based on text description |
| Batch Inference API | gRPC | Asynchronous processing of large request batches |
| Text Classification API | gRPCREST | Classification, moderation, and topic spotting |
| Embeddings API | gRPCREST | Text vectorization |
| Files API | gRPCREST | Loading and storing files and data |
| Dataset API | gRPC | Training dataset management |
| Tuning API | gRPC | Domain-based model fine-tuning |
| MCP Gateway API | gRPCREST | Creating, listing, and deleting MCP servers |