🤖 Creation of CHT-Agent / Multi-Agent Squad

Hi everyone,

I’m excited to kick off the CHT-Agent / Multi-Agent Squad!

Our work focuses on developing a Hierarchical Multi-Agent System designed to empower community health systems built with the CHT. The goal is to build a collaborative environment where specialized agents covering research, development, and QA, work together to automate workflows and improve the CHT development process.

You can dive into the technical design here: CHT Hierarchical Multi-Agent System PoC.

:spiral_calendar: Coordination & Weekly Sync

I’ll be leading coordination for this squad to help keep our progress moving.

One of our first priorities is finalizing the Research Supervisor for release and sharing that first feature during a community call. To stay aligned, I’d like to establish a weekly sync call.

Please reply with your timezone and a preferred 60-minute window so we can find a recurring time that works for everyone interested in participating.

:light_bulb: Ways to Contribute

We want to ensure anyone interested can support the squad, regardless of whether you have a Claude subscription. There are several ways to contribute:

  • System Development: Working directly on the agent architecture and multi-agent coordination.
  • Issue Categorization: Helping categorize existing GitHub issues into domain categories (webapp, api, shared-libs, etc.). This is crucial for building context and improving agent accuracy.
  • Testing & Review: Validating agent outputs, testing the Research Supervisor, and providing feedback on system logic.
  • Documentation: Improving agent prompts, context files, and system documentation.

:waving_hand: Join the Squad

If you’d like to be involved, please reply to this thread and let us know:

  1. What part of the vision interests you most?
  2. Your availability for the weekly sync call (timezone + preferred time window).
  3. Which areas you’d like to focus on (development, categorization, testing, documentation, etc.).

Looking forward to building this together!

6 Likes

:tada:
This is super exciting!

What part of the vision interests you most?

All of them :heart_with_ribbon:

Your availability for the weekly sync call (timezone + preferred time window).

I’ll be able to join a meeting sync every other week / bi-monthly. Preferred day(s) are Tuesday or Wednesday. Since my time zone will likely be an outlier (MST): I’m flexible participating during my late evenings, like 11pm MST.

2 Likes

I’d sure like to be involved in this.

What part of the vision interests you most?

The interesting frontier of AI agents.

Your availability for the weekly sync call

Tuesday or Wednesday also works for me 5PM - 8PM EAT.

Which areas you’d like to focus on:

I am flexible with development, categorization, testing,

3 Likes

Hi everyone,

Looking at the preferences shared so far, it seems Wednesdays are a great overlap day for us. To help bridge the gap between MST, EAT, and NPT, I wanted to propose two specific slots on Wednesdays to see which the group prefers:

  • Option 1: 7:45 PM NPT / 7:00 AM MST / 5:00 PM EAT
  • Option 2: 8:45 PM NPT / 8:00 AM MST / 6:00 PM EAT

Option 1 lands right at the start of Ben’s preferred window, though it’s quite an early start for Hareet. Option 2 moves us slightly later into the evening for those of us in EAT/NPT but might be a bit more manageable for the MST morning.

Does either of these work better for the squad? Looking forward to getting this started!

@sugat I was reading the Multi Agent POC, was wondering do we need this 1st checkpoint HUMAN VALIDATION CHECKPOINT #1?

The timeline looks good to me, will be available for commit

2 Likes

Exciting! I’d like to be part of this; please add me to the squad.

  1. What part of the vision interests you most? - AI agents.

  2. Your availability for the weekly sync call (timezone + preferred time window). Mon/Wed 6 PM onward, EAT

  3. Which areas you’d like to focus on (development, categorization, testing, documentation, etc.). - Development

2 Likes

Yes, you’d need to validate the research and plan before implementation starts. Although the research output should improve over time as we build better context, it’s still important to review it at this checkpoint. This prevents wasted development effort if the research agent misunderstands the issue or suggests the wrong approach.

Thanks for all the responses!

It looks like the best overlap for the majority is Wednesdays at 8:45 PM NPT / 8:00 AM MST / 6:00 PM EAT.

Please react to this message (:+1:) if this works for you, or let me know if you have any major conflicts with this slot.

If there are no objections, I’ll go ahead tomorrow and schedule this as a bi-weekly meeting starting next week. Looking forward to it!

Ahhh so we are basically building a good dataset by using this framework? Then further we are going to take it to prod without human in the loop, is it?

Yes and no. Each task the system solves creates context that improves future outputs, so the agents get better over time. However, we’ll likely always keep the human checkpoints. They’re a core part of the architecture, not just training wheels. Even in production, you’d want to review research before committing development resources, and validate implementation before merging code.

1 Like

I’ve officially created a recurring bi-weekly meeting in the CHT calendar for Wednesdays at 8:45 PM NPT / 8:00 AM MST / 6:00 PM EAT. Please add it to your calendar as well to receive notifications.

Looking forward to the kick-off next week!

CHT Multi-agent System Squad
Google Meet joining info
Video call link: https://meet.google.com/dfv-zeoa-swp
Or dial: <U+202A>(US) +1 305-912-4546<U+202C> PIN: <U+202A>521 014 686<U+202C>#
More phone numbers: https://tel.meet/dfv-zeoa-swp?pin=8992537402696
3 Likes

Could you add me to the meeting invite? I amnt able to make an event with this particular invite link

@shashi_sah, I have added you to the meeting invite.

2 Likes

Dear Community, appologies for the late. I am available for the further meeting. Kindly mention me over the list.

Looking forward.

2 Likes