This is my first attempt at working with LLMs to create Intelligent Digital Workers.

Eventually this project will implement the ideas discussed here

For now, this simple example just connects to a running Ollama server and streams it's response to a hard-coded question.

ollama serve

Ollama requires a model on your system. To install the llama3 model, for instance, run the following command:

ollama run llama3

python main.py

As of now, some functionality has been shown. But it is very slow (~3-4 minutes to see activity), and on top of that there is an error that is also encountered

I encountered an error while trying to use the tool. This was the error: DuckDuckGoSearchRun._run()
got an unexpected keyword argument 'q'.
Tool duckduckgo_search accepts these inputs: A wrapper around DuckDuckGo Search. Useful for when
you need to answer questions about current events. Input should be a search query.

There is a GitHub issue tracking this same error, but it doesn't appear to have been triaged as of yet.

One thing to note, is that depending on which model is selected for the research_agent role, it might bypass the error after a few tries. This was seen when using the solar model.

Will look into trying other tools to replace DuckDuckGo search.