Great progress! But the project isn't working right now, is it?

[shanmu-raoyunfei]

There is a lot of progress on the project right now, and you can see a lot of improvements. At present, the online version does not work properly. For example, when viewing the python code, it will give an error, and it can not reply to the problems such as running (and I do not see the location of the openai key). It does not work for the time being, right?

[hughlv]

This is expected behavior. You can take a look at the few templates https://studio.agentok.ai/marketplace I published recently for reference. (This should be mentioned in document, or as part of better UX. Not a priority to me for now.)

As a complete workflow, the first node should be Initializer, the second one can be any ConversableAgent, or User node, otherwise the workflow cannot be turned into a workable python program.

However, even though I think the basic structure should be fine now, I'm still considering the value of AutoGen, later on will iterate based on actual scenarios. If you're a programmer, any contributions are welcome.

[hughlv]

Today I did a simple patch to avoid your situation: When creating a new project, a default working workflow will be provided as sample for further editing.

[shanmu-raoyunfei]

Many thanks

[shanmu-raoyunfei]

Why does the demo in any of the markets I selected not work? In addition, there is no deletion in the project I fork out?

[hughlv]

Why does the demo in any of the markets I selected not work? In addition, there is no deletion in the project I fork out?

No worries. I'm currently spending all my spare time doing a huge architecture refactoring. Will be back soon.

[shanmu-raoyunfei]

Great, because it's a great project, so keep an eye on it and think about how it applies to production. Thank you for your efforts

[hughlv]

Great, because it's a great project, so keep an eye on it and think about how it applies to production. Thank you for your efforts

Understood. I'm happy you like it. And it's the reason why I should spend more time on the improvements.

[shanmu-raoyunfei]

When will this great project be updated, thanks

[hughlv]

When will this great project be updated, thanks

I'm facing some difficulties integrating the RAG feature of autogen. Will try to catch up as soon as I get some spare time.

[hughlv]

When will this great project be updated, thanks

@shanmu-raoyunfei I saw your Feature Request on AutoGen about RAG microsoft/autogen#1387. Have you managed to get RAG agent working? I'm stuck by their docs and examples, because all the example does not seems to solve real problems, just pass in some docs and vectorize, this is definitely no the real case. What I'm expecting is to use an existing external knowledge base directly.

[shanmu-raoyunfei]

To be honest, all RAGs are currently mostly vectorized and used later. I understand what you mean, but it would be acceptable for us to vectorize the knowledge base when we configure it for the agent. There is an autogen open source project that can use data sources directly RAG, you can refer to below. The name of this project is: agentcloud

[hughlv]

To be honest, all RAGs are currently mostly vectorized and used later. I understand what you mean, but it would be acceptable for us to vectorize the knowledge base when we configure it for the agent. There is an autogen open source project that can use data sources directly RAG, you can refer to below. The name of this project is: agentcloud

I know this project. will check out in detail later. thx.

[shanmu-raoyunfei]

Very happy to see the project has made great progress, more and more look forward to!

[shanmu-raoyunfei]

In addition, is our architecture BaaS, and does it support calling workflow through external API?

[hughlv]

In addition, is our architecture BaaS, and does it support calling workflow through external API?

Yes it supports API by design. you can take a look at https://api.agentok.ai/v1/docs, which contains all the exposed APIs and you can test out with API key.