Skip to main content

Welcome to Compressa Documentation!

Compressa offers a ready-to-use set of powerful models with Russian language support for custom development of LLM + RAG assistants based on Langchain. Using either our ready-made APIs or deploying a copy of the platform on your own servers, you can create the AI assistant you need for your specific business task.

Some examples of possible solutions:​

  • QA-bot for your set of documents across various departments and tasks
  • Smart and accurate search for employees or clients
  • Langchain agent that executes multiple steps to complete a task
  • Analysis and segmentation of text feedback from clients

Compressa provides optimized LLM, Embedding, and Rerank models for the Russian language. However, thanks to native integration with Langchain, you can assemble hundreds of different AI agents.

When using the on-premise format, the LLM can also be quickly and inexpensively fine-tuned for your task using LoRA adapters.

How to use our models?​

You can use our models and tools in several ways:

  • Python SDK (Langchain): We provide a Python SDK to simplify access to our models and tools from your Python code.
  • REST API: Our SDK uses a public REST API that you can call directly from your code in any language.
  • OpenAI Compatible: Our APIs are compatible with OpenAI, so you can simply use their library with our models.

Who can use our platform?​

The current version of the platform is best suited for developers who are already familiar with Langchain and want to use ready-made models from Compressa. If you haven't worked with this framework yet, Langchain contains many ready-made templates and libraries that you can take and use with our models. We have also prepared a set of guides for quick start and some common scenarios.

If you need other/additional platform functionality, please contact us in the Telegram support chat.

Quick Start​

Try our models in practice with one of the quickstart guides:

⁃ Quick Start with Python SDK (Langchain)
⁃ Quick Start with REST API
⁃ Transition from OpenAI