An early experiment to replicate AutoGen's examples with ollama local models.
Install AutoGen with: pip install pyautogen
You'll most likely need to pip install yfinance matplotlib
When generated code fails with module not found errors, give it some human assitant and pip install things for it.
Note: There are other ways to execute code besides directly running using LocalExecutor and manually pip installing things. However, I have mixed results using Docker and LocalExecutor with VirtualEnvironment.
Your millage may vary.
Mostly, all the examples are based from AutoGen's examples
Give a simple prompt and use executor to repeatedly try until it gets the result
python auto_simple.py
This example is more akin to Software Developers usual tasks. Where we're in the loop and continuously give feedback to the agents.
python coding_task.py
The most succesful experiment is having multiple agents start doing research on their own and then have them communicate with each other. While having a Planner to reflect and look at the output.
python research.py
pip install autogenstudio
autogenstudio ui --port 8081