top | item 36600966

Show HN: Lemon AI – open-source Zapier NLA to empower agents

13 points| brockmeier | 2 years ago |github.com | reply

Hi HN, excited to share this open source Zapier NLA alternative I’ve been working on.

A few weeks ago I found myself struggling to automate a relatively straightforward internal workflow using an open source LLM (update Hubspot based on Airtable entry and personal information coming from web sources, and send out a summary via Slack). There was no way to interact with these tools in a robust way so I decided to build my own connectors.

While the existing connectors out there are helpful to reduce the risk of hallucinations by reading data from sources, agents only become truly powerful when they are able to reliably perform workflows for us. For this, my view is that a secure and robust way to execute write operations is needed.

Lemon AI allows for both read and write operations in various tools - even if you work with open source LLMs. It can be used in combination with Langchain and allows for easy integration with your existing agents.

I would love your feedback on what can be improved!

7 comments

order
[+] johnartss|2 years ago|reply
Super interesting, we are facing similar challenges. Let me know if you want some feedback over a call.
[+] arnestrickmann|2 years ago|reply
This is really cool. Recently found Foundry on Discord, how is this different?
[+] brockmeier|2 years ago|reply
Foundry is doing a nice job bringing functions in LLM apps to life. That's something I am also working on with Lemon AI functions. Besides that, to me it seems that it’s focused on read operations for web-based services. What I’m building is focused on allowing for read and write operations, both for internal tooling and web services to allow agents to take actions based on instructions. I’m essentially focussing on helping Langchain users to establish end-to-end workflow automation as quickly as possible. Going forward permission workflows will be covered as well so assistants can be team / user specific, something that I believe isn’t solved well by Zapier NLA right now.