Integrate AgentOptimizer

[skzhang1]

Why are these changes needed?

AgentOptimizer, a new class for training AutoGen agents in the era of LLMs as a service. AgentOptimizer is able to prompt autogen agents to iteratively optimize their function/skills according to historical conversation and performance.

Currently, it only supports training one Assistant-UserProxy agent pair and is subject to be extended to a large scenario in the future.

Why we need to separate agents and AgentOptimizer?

We dedicate to give users more room for customization in their own scenarios. The AgentOptimizer is solely for generating functions/skills, how/which/when to register should be decided by users.

TODO:

• [x] Determine API for Agentoptimizer, which is the basic for
• [x] Add notebooks to introduce the concept of agent training
• [x] Add test and corresponding documentation

Currently. this PR is a draft intended for facilitating discussions related to API interfaces. Any suggestions especially on the API design are appreciated.

Related issue number

N/A

Checks

• [ ] I've included any doc changes needed for https://microsoft.github.io/autogen/. See https://microsoft.github.io/autogen/docs/Contribute#documentation to build and test documentation locally.
• [ ] I've added tests (if relevant) corresponding to the changes introduced in this PR.
• [ ] I've made sure all auto checks have passed.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 0% with 149 lines in your changes are missing coverage. Please review.

Project coverage is 47.63%. Comparing base (59a7790) to head (b8192ee). Report is 1 commits behind head on main.

Files	Patch %	Lines
autogen/agentchat/contrib/agent_optimizer.py	0.00%	148 Missing and 1 partial :warning:

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #1767       +/-   ##
===========================================
+ Coverage   36.73%   47.63%   +10.89%     
===========================================
  Files          69       70        +1     
  Lines        7134     7283      +149     
  Branches     1557     1717      +160     
===========================================
+ Hits         2621     3469      +848     
+ Misses       4280     3516      -764     
- Partials      233      298       +65     

Flag	Coverage Δ
unittests	47.49% <0.00%> (+10.76%)	:arrow_up:

Flags with carried forward coverage won't be shown. Click here to find out more.

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[skzhang1]

@qingyun-wu @sonichi This PR is ready to merge.

[qingyun-wu]

The code looks good. But not sure if this should live in "extensions".

[JieyuZ2]

The code looks good. But not sure if this should live in "extensions".

we can move to contrib?

[skzhang1]

@JieyuZ2 agree, contrib is better. @qingyun-wu what do you think?

[skzhang1]

The code looks good. But not sure if this should live in "extensions".

done. move to contrib

[ekzhu]

Would love to have a topic page under /docs/topics/ on research related topics like continuous improvement.