Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Feature Request]: Autogen workflow deployment examples using LLMOps practices like CI/CD, GitOps #44

Open
johnnieskywalker opened this issue Nov 19, 2024 · 0 comments
Labels
enhancement New feature or request

Comments

@johnnieskywalker
Copy link

Is your feature request related to a problem? Please describe.

Deploy autogen workflows in LLMOps manner with CI/CD pipeline triggered by github action and then available as endpoints to stream responses from workflows as websocket or server side events. So far I did it by deploying autogen workflows inside prompt flow and to deploy prompt flow itself I used this LLMOps examples repository. Then as the last node of flow a python tool that uses yield to stream the output from flow, and wrapping it into managed api endpoint in azure to configure cors etc. We know it's overcomplicated, it would speed up adoption of AG2 in prod environments if there were some tutorial how to perform LLMOps.the

Describe the solution you'd like

Tutorial, or example repository that would show example how to create LLMOps cycle to deploy Autogen2 workflows to various environments (like dev, prod etc.). Ideally, it should contain examples:

  1. How to stream ag2 workflow input in real-time chunks using for example WebSocket or server-side events.
  2. What is the recommended infrastructure to run ag2 workflow in cloud
  3. GH action or other CI/CD pipeline that demonstrates how to build ag2 workflow when pull request is merged
  4. Observability patterns - how to monitor ag2 agents on various environments browse logs etc.
  5. Recommendations on e2e testing ag2 workflows, for example using MLFlow to see if agents don't hallucinate, use another llm to see if output makes sense or some statistical methods to make assertions on outputs.

I know it's wide topic, but it would help mass adoption of ag2 and make it used on production by many teams.

Additional context

Ag2 itself is wonderful tool for building MAS (multi agent systems) but it's true potential lies in connecting it with web apps, mobile apps etc. on production with great UI/UX. This would allow ag2 to solve real-world problems and make it widely adopted both in non-profit organizations and for-profit companies. If we want ag2 to be the next big thing in 2025 we need to have some established patterns on how to test it and use in production environments.

@johnnieskywalker johnnieskywalker added the enhancement New feature or request label Nov 19, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

1 participant