deepseek

This is a module that uses AI specifically self-hosted using LM Studio or Ollama and free models like Deepseek. It can add into ckeditor5 to generate content, create chatbots that allow direct exchange, AI agents, with RAG embeding and Model Context Protocol support to be able to scan the content of articles or files to enrich content

Require

Your database must support VECTOR type.
It must have MySQL version 9+ or Mariadb 11.7+ for RAG
You can use AI providers like Deepseek, OpenAI, or Gemini.
If you want to host your server yourself, you can install LM Studio on your server.

How to use

• You must install and run LM Studio or Ollama on the server. Load the necessary free models, I highly recommend you use deepseek 7B or Gemma 3 which are very light for weak servers. The nomic-embed-text or mxbai-embed-large models are needed if you want to use embedding
• After installing the module, go to the settings module, you can configure the url and api key (for gemini). You can define the types of prompts used for ckeditor.
• You can add the ai chat block or go to the /chat path to be able to chat with AI.
  
• You can use VBO to create a view and select the entities you want to embed into the document so that the AI ​​chat can be smarter.
• You can define functions and permissions so you can ask AI to do something when chatting. (MCP vs function calling is new so this feature will be updated latter).
• If you want RAG your image, document, pdf.... you can do with module
  OCR File
• Chat inteface is using bootstrap 5, you can restyle template with your custom theme, fully support bootstrap 5 admin

This is a module for personal research on self-hosted solutions so you should be sure about the hardware before installing.

If you find this module valuable, consider supporting the developer by buying me a ☕