Show HN: Lemon Agent – Plan and Solve agent for workflow automation (github.com)

7 points by brockmeier 2y ago ↗ HN

I built this Plan and Solve (PS) agent because I struggled to build stable sequential tasks with LangChain and GPT3.5 turbo. After many failed attempts with LangChain's PS agent, I decided to build my own.

I built the agent in a way that it acts as a workflow automation co-pilot rather than trying to play AGI. For example, the user has to tell the Lemon Agent which task needs to be solved and whether the suggested workflow automation is the right one for the task. I decided to go with a PS architecture since I consider the separation of responsibility a great way to increase the accuracy of agents, which is one of the main reasons why I built this agent. I am actively thinking about adding a workflow automation-specific third agent that helps with input validation (so kind of a ‘PVS’ agent) to make the input parameter provision in each workflow step more accurate.

I’d love to hear your thoughts and get some initial feedback!

2 comments

[ 3.0 ms ] story [ 11.9 ms ] thread

hainghiem375 2y ago ↗

Will this compete with Zapier or other low-code/no-code platforms?

brockmeier 2y ago ↗

I think Zapier's NLA solution goes in a similar direction and as far as I know, they are running their model on the server side to adjust workflow input and output of workflows to the unique needs of an LLM model. They are doing an amazing job there and I am also currently working on improving this (e.g. adjusting to a model's token limit).

Besides that, Lemon Agent focuses on providing an out-of-the-box agent that is great at solving tasks with workflow automation. If you want to use Zapier NLA you have to rely on generalistic agents like LangChain's ReAct Agent or the Plan and Execution Agent or you have to build your own agent. In any case, you have to twist the agent a lot to reach a certain degree of accuracy and reliability.