Tulip Agent -- Enabling LLM-Based Agents to Solve Tasks Using Large Tool Libraries

TL;DR

The tulip agent architecture enables large language model-based agents to efficiently utilize extensive tool libraries through recursive search, reducing inference costs and supporting adaptability across various domains.

Contribution

It introduces a novel architecture that allows LLM agents to access large, extensible tool libraries without encoding all tools in the prompt, improving efficiency and scalability.

Findings

Reduces inference costs significantly.

Supports large and extensible tool libraries.

Demonstrates generalizability in robotics.

Abstract

We introduce tulip agent, an architecture for autonomous LLM-based agents with Create, Read, Update, and Delete access to a tool library containing a potentially large number of tools. In contrast to state-of-the-art implementations, tulip agent does not encode the descriptions of all available tools in the system prompt, which counts against the model's context window, or embed the entire prompt for retrieving suitable tools. Instead, the tulip agent can recursively search for suitable tools in its extensible tool library, implemented exemplarily as a vector store. The tulip agent architecture significantly reduces inference costs, allows using even large tool libraries, and enables the agent to adapt and extend its set of tools. We evaluate the architecture with several ablation studies in a mathematics context and demonstrate its generalizability with an application to robotics. A reference implementation and the benchmark are available at github.com/HRI-EU/tulip_agent.