feat: support multiple workbenches in assistant agent

[Griffin Bassman (bassmang)]

resolves: #6456

[Alex-Wenner-FHR (Alex-Wenner-FHR)]

I pip installed this branch because this is a feature I am in need of, and did notice the param in the _call_llm is incorrect, expects workbenches but gets passed workbench and causes an error. Griffin Bassman (@bassmang)

[Griffin Bassman (bassmang)]

I pip installed this branch because this is a feature I am in need of, and did notice the param in the _call_llm is incorrect, expects workbenches but gets passed workbench and causes an error. Griffin Bassman (@bassmang)

Alex-Wenner-FHR (@Alex-Wenner-FHR) fixed this issue
Jack Gerrits (@jackgerrits) I changed the naming of everything in assistant_agent to workbench. This change is still breaking though since it saves workbench as a list when serializing and deserializing

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 79.54%. Comparing base (6cadc7d) to head (286f1f0).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #6529   +/-   ##
=======================================
  Coverage   79.54%   79.54%           
=======================================
  Files         225      225           
  Lines       16662    16669    +7     
=======================================
+ Hits        13254    13260    +6     
- Misses       3408     3409    +1     

Flag	Coverage Δ
unittests	79.54% <100.00%> (+<0.01%)	⬆️

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

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Victor Dibia (victordibia)]

Looks good. I tested with the following.
Thanks Griffin Bassman (@bassmang)

from autogen_ext.models.openai import OpenAIChatCompletionClient
from autogen_ext.tools.mcp import StdioServerParams, McpWorkbench
from autogen_agentchat.agents import AssistantAgent
from autogen_core import CancellationToken
from autogen_agentchat.ui import Console 
 
# Setup server params for local filesystem access
fetch_mcp_server = StdioServerParams(command="uv", args=["tool","run","mcp-server-fetch"], read_timeout_seconds=100  )
git_mcp_server = StdioServerParams(command="uv", args=["tool","run","mcp-server-git"], read_timeout_seconds=100)
    
fetch_wb = McpWorkbench(server_params=fetch_mcp_server)
git_wb = McpWorkbench(server_params=git_mcp_server)

# Create an agent that can use the fetch tool.
model_client = OpenAIChatCompletionClient(model="gpt-4o")
agent = AssistantAgent(name="fetcher", model_client=model_client, workbench=[fetch_wb, git_wb], reflect_on_tool_use=True)

 
await Console(agent.run_stream(task="Find and summarize the latest article on https://www.pragmaticengineer.com/ and  What is the git status of https://github.com/microsoft/autogen ", cancellation_token=CancellationToken()))

Results look good

[shabd01]

Hi Griffin Bassman (@bassmang), Eric Zhu (@ekzhu)

Does this feature supports Cross-workbench session sharing?

For example Can a browser session created from Playwright workbench can be accessed by a StaticWorkbench?