0.6.0-alpha.2

• Add reasoning model support in classic chat page.
• Disruptive change: Starting from v0.6.0, the schema of the datasheet that records classic chat has been modified and is not compatible with the local storage of the previous version, so please make a backup.

What's Changed

• Upgrade to autogen v0.4 by @Wannabeasmartguy in #17

Full Changelog: v0.5.2.2...0.6.0-alpha.2

v0.5.2.2

Updated to AutoGen v0.4

I've been following the development of AutoGen for a while now and have written a test application to some extent. Now that AutoGen has entered a new phase, I think it's ready to be stabilized! So I've spent some time adapting it and added a new page for configuring Agent Team.

New Agent Chat Page

With AutoGen v0.4, you can easily unlock the potential of LLM. In Agent Chat page, you can check how an Agent Team with multiple Agents operates internally.

Agent Team Template Manage Page

We have two sub tabs in this page, the first is a agent team library:

You can check and modify all team templates here.

The second tab is used to create a new agent team, and for now we just support reflection pattern:

Creating an Agent Team is a two-step process:

1. Select a model;
2. Fill in the information on the form and submit it.

Now, the selected model will be used for all Agents, and customization of models for individual Agents may be supported in the future.

What's Changed

• Native RAG, local storage for chat conversations and configurations, and more beautiful interactive pages by @Wannabeasmartguy in #3
• docs: add Japanese README file by @eltociear in #4
• Bump pillow from 10.0.0 to 10.3.0 by @dependabot in #5
• fix bug by @Wannabeasmartguy in #6
• Refactor knowledge base management by @Wannabeasmartguy in #7
• Refactor knowledge base management by @Wannabeasmartguy in #8
• Reorganize directory structure by @Wannabeasmartguy in #10
• Fix Bug by @Wannabeasmartguy in #11
• Decouple dependencies from AutoGen v0.2; Bug fixes. by @Wannabeasmartguy in #13
• Fix run_id index bounds and enhance dialog retrieval by @Wannabeasmartguy in #14
• Resolve bugs in dialog management by @Wannabeasmartguy in #15
• More robust dialog state switching by @Wannabeasmartguy in #16

New Contributors

• @eltociear made their first contribution in #4
• @dependabot made their first contribution in #5

Full Changelog: v0.2.1...v0.5.2.2

v0.2.1

Now, I am finally very happy to release the first Release of RAGENT.

First of all, RAGENT has a very clean and intuitive interface, which is not easy for an inexperienced developer. Happily, the current interface is great, and when I look at it as a user, I'm very happy with it, it's my current go-to webui, and I can do almost all of my simple model configuration work from within the page.

In this version, I focused on rewriting the RAG-related code, ditching the original code that used Langchain and trying to fully customize the functionality, and I'm happy that the modules still work great, even on my GPU-less RYZEN PC. I hope you enjoy it too!

Here are my thoughts on features that may be added in future releases that are in the pipeline:

• Multimodal model support (OpenAI Compatible);
• More embedded model support;
• Intermediate process display for Function Call;

If you have a good idea you'd like to share with me, feel free to raise an issue!

Full Changelog: v0.2.0...v0.2.1