[Feature Request]: A more modular way of registering a function with an agent.

[tyler-suard-parker]

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

I like Autogen better than Semantic Kernel, but one thing that Semantic Kernel does really well is making its functions available to agents in a very modular way. For example, in autogen we need to create a long dictionary of functions, and in order to add a function to our project and to an agent, we have to modify that dictionary. Here is an example:

llm_config = {
   "functions": [
       {
           "name": "define_function",
           "description": "Define a function to add to the context of the conversation. Necessary Python packages must be declared. Once defined, the assistant may decide to use this function, respond with a normal message.",
           "parameters": {
               "type": "object",
               "properties": {
                   "name": {
                       "type": "string",
                       "description": "The name of the function to define.",
                   },
                   "description": {
                       "type": "string",
                       "description": "A short description of the function.",
                   },
                   "arguments": {
                       "type": "string",
                       "description": 'JSON schema of arguments encoded as a string. For example: { "url": { "type": "string", "description": "The URL", }}',
                   },
                   "packages": {
                       "type": "string",
                       "description": "A list of package names imported by the function, and that need to be installed with pip prior to invoking the function. This solves ModuleNotFoundError.",
                   },
                   "code": {
                       "type": "string",
                       "description": "The implementation in Python. Do not include the function declaration.",
                   },
               },
               "required": ["name", "description", "arguments", "packages", "code"],
           },
       },
   ],
   "config_list": config_list,
   "request_timeout": 120,
}

Semantic Kernel, on the other hand, stores each function in a separate file as a "plugin". In order to make the agent aware of the function, they simply call

kernel.import_skill(skill_name).  

It would be nice to have that functionality in Autogen, as it would simplify the process of adding new functions.

Describe the solution you'd like

from functions import web_search_function

assistant_agent.add_function(web_search_function)

The assistant agent can now use that function, and is aware of its description and required inputs and outputs.

Additional context

No response