SWP26

LLM-based agents (Software project; WS 26)

Here are the slides from the first meeting

Interactive agents, which autonomously complete tasks through interactions with their environment, can assist humans in various domains of their daily lives. LLM-based agents are AI system that uses a Large Language Model (LLM) to solve a complex task, including task understanding, solution planning, and executing the task by interacting with external tools, APIs, databases, etc.

The goal of this software project is to identify a relevant use case for such software and implement and evaluate an agent for complex multi-step tasks. Our implementation will be based on the MCP (Model Context Protocol) protocol, an open-source protocol for connecting LLMs to external components, and the LangChain/LangGraph framework for implementing tge agent workflow.

Information

Instructor: Mareike Hartmann (mareikeh@coli.uni-saarland.de)
Time & Place: Mondays, 12:15-13:45, C7.2, Seminarraum -1.05
Course structure:

• Introduction, brain storming and finding a topic during the first 3 weeks. Then, students will work in small groups of 3-4 people.
• During the semester, weekly meetings with the whole course in the Monday slot for progress presentations.
• Weekly 30mins meetings with me and your group to discuss implementations, problems, etc.
• Final deadline for handing in the project: August 31

Registration

• Please attend the kick-off meeting on Monday, April 13, 12:15 in C7.2, Seminarraum -1.05
• send me an email (mareikeh@coli.uni-saarland.de) until end of day (April 13), including:
  • name
  • study programme
  • why you want/need to take this course
  • why are you interested in LLM-based agents
  • programming experience
  • experience with LLM-based agents in professional or private life (if are you using them and how)