Openagents: An Open Platform For Language Agents In The Wild

We present OpenAgents, an open platform for using and hosting language agents in the wild of everyday life. OpenAgents includes three agents: (1) Data Agent for data analysis with Python/SQL and data tools; (2) Plugins Agent with 200+ daily API tools; (3) Web Agent for autonomous web browsing. OpenAgents enables general users to interact with agent functionalities through a web user interface optimized for swift responses and common failures while offering developers and researchers a seamless deployment experience on local setups, providing a foundation for crafting innovative language agents and facilitating real-world evaluations.

During building OpenAgents, we first underscore the significance of effectively specifying application requirements via LLM prompting, a process that often requires crafting instructions that cater to backend logic, enhance output aesthetics, and safeguard against adversarial inputs. Our findings indicate that the build-up of such instructions can, at times, be substantial, posing challenges in terms of token limitations and context handling for the LLMs. Additionally, for effective realworld deployment, agent models must not only exhibit high performance but also be able to handle real-time, interactive scenarios, such as streaming, to provide an optimal user experience. Furthermore, our exploration reveals that current research often gravitates towards idealized performance metrics, sometimes sidelining critical real-world considerations, such as the trade-offs between system responsiveness and accuracy, and the nuanced complexities introduced when application-based failures arise, potentially obfuscating the true capabilities of the LLMs.