Lindy Community Icon

Home
Events

00 Announcements
01 General
02 Support
03 Bugs
04 Requests and Feedback
05 Share Your Lindy
06 Intros

Community Guidelines

Troubleshooting Guide

Powered by Tightknit

Creating a Manager Agent for Parallel Sub-Agent Execution in Lindy | Lindy Community

Creating a Manager Agent for Parallel Sub-Agent Execution in Lindy

·

·

Hi Support Team, my first post here. Exploring Lindy quite a bit and created some nice flows. Love the AI Agent focused approach (verses code all over). I am using it for basic app testing, using the computer, login and then press some buttons. Quite an evolution from what we have been doing for the last 30 years. I am able to get Agents working great.. and it does exactly what is needed. I am running in to challenges with creating a ‘manager’ agent that can start and run sub agents ‘in parallel’. I have tried conditions with more than one parallel exit(does not work) I have loops, one got stuck and ate all my credits.. so bad. Very brute force and not easy to scale and add additional. I am looking for the correct mechanism to enable a manager agent to start sub agents .. in parallel, and all run and then when all are done I generate a summary report. Can see how that would be useful and can have a suite of test cases running and then send a summary. Can someone share the right mechanism for this? Or even better a working example of pure agent to agent parallel invocation ? Really appreciate any insight here.

14 comments

· Sorted by Oldest

Alon J.
·
·
Hey Ray! Thanks for reaching out! Happy to take a look at your workflow and provide some feedback. Would love to see what happened exactly and how they got stuck. Do you mind sharing your Task URLs, so we can take a look? How to share your Lindy task URL: 1. Click your Lindy agent → click “tasks” 2. Find the task you need help with 3. Click the specific task (you should see the full execution) 4. Copy that page URL from your browser 5. Paste the URL in your support request
Ray
·
·
Alon J. here is one of them: https://chat.lindy.ai/mancomm/lindy/manageragent-68f96d805d0c3253dfaa015b/editor
Ray
·
·
Whats the proper way to create a manager agent that can kick off N number of parallel sub agents. Is there a special action that can do that?
Ray
·
·
I also have this loop one: https://chat.lindy.ai/mancomm/lindy/multi-agent-parallel-test-coordinator-68f990a76daa14d3f1927dd7/editor
Ray
·
·
but I am just asking whats the right way to do that.. just a simple agent that says 'hi' ,.. any example?
Alon J.
·
·
Hey Ray! I'll send you a DM. Seems like you have a bunch of questions - Happy to clarify!
Ray
·
·
Only one question just asking it different ways trying to get the same result 🙂 Happy to DM!!
Ross
·
·
I would go with webhooks for invoking agent to agent calls.
Ray
·
·
Ross Lindy has agent to agent communication built in already. They are called Agent messages and reply to agent.. thats not the issue. The issue is starting more than one in parallel. Loops seem to be the only way today
Ray
·
·
👍
Ross
·
·
yes, any parallelization using Lindy constructs would require parallel. My answer was if you wanted agents in different lindy accounts to work. I don't think agent messages cross the 'account' barrier. Or, you could do it with code. Call computer use, create webhook messages there, and then the agent will suddenly get say 10 triggers for 10 messages, thereby creating 10 tasks
Ray
·
·
Ok thats fair. I was not on that use case and I use the webhook to notify slack.. its a very nice capability. I suggested to Alon J. that Lindy put in the pipeline a more parallel out put connection from a special task.. so all outputs are invoked at once.. verses just one. That would remove the complexity of loops. Loops are a very very messy way to enable parallel processing.
Ross
·
·
I agree, and loops are fraught with danger 'eating credits'
Ray
·
·
I lost 800+ last night because the loop got stuck .. quite frustrating. My own fault for not putting in guards .. but still.. 800 credits is 800 credits..