astra-langchain4j: Experiences Combining LLMs and Agent Programming
Rem Collier, Katharine Beaumont, Andrei Ciortea
TL;DR
This paper addresses integrating large language models into traditional agent programming (ASTRA) to explore agentic AI workflows. It presents astra-langchain4j, a LangChain4J-based library enabling LLM calls, prompt templating, and BeliefRAG for retrieval-augmented generation within ASTRA. Three example systems—Travel Planner, Tic-Tac-Toe, Towerworld—illustrate how LLMs can participate in multi-agent coordination, decision making, and planning via prompts and composite templates. The experiences reveal that while integration is straightforward and supports agentic workflows, LLMs exhibit limited contextual reasoning and consistency, underscoring the need for careful prompting and plan-design strategies. The work demonstrates practical steps toward leveraging LLMs within established MAS toolkits and highlights trade-offs between single-agent versus multi-agent designs.
Abstract
Given the emergence of Generative AI over the last two years and the increasing focus on Agentic AI as a form of Multi-Agent System it is important to explore both how such technologies can impact the use of traditional Agent Toolkits and how the wealth of experience encapsulated in those toolkits can influence the design of the new agentic platforms. This paper presents an overview of our experience developing a prototype large language model (LLM) integration for the ASTRA programming language. It presents a brief overview of the toolkit, followed by three example implementations, concluding with a discussion of the experiences garnered through the examples.