How to automatically get data from alfresco and store in vector database like Chromadb

[sanogo]

Hi, I have to build a Chatbot that allows to upload and search files in Alfresco using LLM. I need some help.

I'm currently proposing following architecture bellow.

I want to use Angular in Frontend and Spring Boot in Backend.   Then I use a MCP Server (PythonSDK version) that allows to connect to Alfresco like datasource. So I can upload file and send from Spring Boot through MCP Server to Alfresco. After storing files in Alfresco I have to vectorize and store in a vector database like Chromadb. Which would allow me to do a search from the expression provided from the frontend.

I use Ollama on purpose to use LLM in local. Also I use langchain with Spring Boot on purpose to implement LLM Api. 

I don't know yet how to automatically store files from Alfresco to ChromaDB.

So I would like to know if it's the right approach.   Because the main purpose is to search and find by an expression contained in a file storing in datasource like Alfresco and ChromaDB.

Thanks !

[angelborroy · Accepted solution]

You have different samples on every missing step:

• Ingestion in Vector DB (like ChromaDB) with Spring AI: https://github.com/aborroy/alfresco-knowledge-enrichment
• Synchronization of Alfresco Contents with Vector DB in https://github.com/aborroy/alfresco-ai-framework

Hope this helps. 

[sanogo]

Hi, thank you for your quit response. This solves my problem.

I will use Synchronization of Alfresco Contents with Vector DB in https://github.com/aborroy/alfresco-ai-framework.

But can I use Angular for UI rather than Alfresco-AI-UI ?

I think to the following architecture:

- Alfresco acts as the Knowledge Base, storing documents.

- alfresco-ai-sync listens to changes in sync folder from the Alfresco Repository and updates the content to the Vector Database through the AI RAG Framework REST API.

- AI-RAG-Framework generates vector representations of the ingested documents and stores them in Elasticsearch or ChromaDB, exposes the REST API. Then, based on request from Angular UI, we search in vector database and send response to Angular UI. We use Ollama which generates vector representations. Also we can send file to store in Alfresco through AI-RAG-Framework.

- Angular for user interface that allows user to upload and store a file in Alfresco by sending through AI-RAG-Framework. Then we can send request through AI-RAG-Framework using langchain4j for searching in Elasticsearch or ChromaDB.