Large language models to support conversational AI with avatar dialogs
Project duration: 4 months
Brief description
In collaboration with students of business informatics, an application is to be developed that extracts data from a document archive using an LLM model (Large Language Model).
Supplement
PTA staff supervised four students in their task of researching and comparing suitable technologies. The focus is on the frameworks LlamaIndex and LangChain and the software AnythingLLM.
Subject description
The developed application runs locally so that sensitive data is not sent to a third-party provider and uses the AnythingLLM software. Furthermore, it has a modular structure so that the components Vectorstore, for storing the data archive, Embedding, for embedding the data in an abstract space, and Language Model can be exchanged as required.