Great work on the new agents release

At a glance

The post is a request from a community member to integrate a custom message queue (MQ) for inter-agent communication, such as RabbitMQ, AWS SQS, or AWS SNS. The comments indicate that integration with RabbitMQ is in progress, and a community member has a working draft for AWS SQS integration. The community members discuss the benefits of using asynchronous clients for the MQ integrations, and one community member shares a link to their initial work on the AWS SQS integration, which they plan to contribute as a pull request.

Useful resources

mmarty mcfly

Great work on the new agents release guys 👏

@Logan M - I've yet to dive into the source for message queue , but could we integrate a custom MQ for the inter-agent communication messages to be posted to? eg. rabbitmq / SQS, SNS or spawn a lambda call?

16 comments

WWhiteFang_Jr

Hi,
Integration with RabbitMQ is in progress: https://github.com/orgs/run-llama/projects/5?pane=issue&itemId=68386242

mmarty mcfly

Good to know! Anything with the AWS stack?

WWhiteFang_Jr

Not fully sure on this, but code abstraction once fully added may help you to setup any message queue.

LLogan M

Glad you like it! Definitely early days but it's been fun building this out.

Yea the hope is the base class well lend itself well to any message queue (probably still some minor kinks of course)

Would love to have Kafka as well, and whatever else 👏

mmarty mcfly

Thanks guys!

mmarty mcfly

@WhiteFang_Jr @Logan M I’ve got a working draft for AWS SQS 💪🏻 but I’ll need some more time to flesh out some more tests.

What would be the best way to contribute, push to a branch so you guys can review?

mmarty mcfly

Also any lessons learned from the previous 2 integrations w/ rabbit and Kafka would go a long way 😄you guys are killer devs, cheers

LLogan M

Amazing! 🤩 :LlamaIndex:

Yea best way to contribute is fork the repo and contribute a PR from your fork 💪

@nerdai has actually been the one killing it on all the recent messsage queue integrations

mmarty mcfly

good to meet you @nerdai , I been reading some of your content on linkedin ✌🏻

aandrei

Hey @marty mcfly! Nice to meet you. Looking forward to your AWS SQS integration!!

aandrei

In terms of lessons from past integrations, I'd say its just that since we're trying to be an async first library, we should make use of async clients. I think for RabbitMQ, I was using sync pika first but making that work with our stuff felt a bit awkward. Switching to aio-pika was better. So hopefully there is a good asyncio equipped client for AWS SQS?

mmarty mcfly

ya, aiobotocore

mmarty mcfly

first stab, this is how far i got, let me know if you have access issues: https://github.com/run-llama/llama-agents/commit/be52f7b7b76eb166c79bd48d896f9293f4974b40#commitcomment-144680990 cheers ✌️

WWhiteFang_Jr

Awesome! You just need to raise a PR from your fork on llama-agents

mmarty mcfly

oops almost forgot - here it is:
https://github.com/run-llama/llama-agents/pull/158
^ fyi @nerdai @Logan M 😄

aandrei

sweet! will take a look at it shortly 🙂

Add a reply

Find answers from the community

Great work on the new agents release