The Do It All basic agent system for Python. It provides a framework for building automated AI agent scripts in Python using a variety of models and API's.
- Creates a path and follows it
- Internet Search
- Image Generation
- RAG system
- Document reading
- Huggingface
- OpenAI
- Google GenAI
pip install git+https://github.com/broadfield-dev/Do_it_All
python doitall/demo.py
pip install git+https://github.com/broadfield-dev/Do_it_All
Save your API keys as the following Environmental Variables to use the models from each.
HF_KEY
(Huggingface API)OPENAI_API_KEY
GEMINI_API_KEY
If you are signed in to Huggingface.co, the demo should work without any modification, using the default model, thanks to Huggingface's free API for all users, up to 1000 requests per day.
from doitall import main
history = []
prompt_in="why is the sky blue?"
print(main.agent(prompt_in,history,mod=2,tok_in="",rand_seed=True,seed=1,max_thought=5,save_mem=False,recall_mem=False,rag_col=False))
The available models can be configured in the doitall/main.py file. The format is as follows:
client_openai=[
{'type':'text','loc':'openai','name':"gpt-4o-mini",'rank':'op','max_tokens':128000,'schema':{'bos':['<|system|>','<|user|>','<|assistant|>'],'eos':'<|im_end|>'},'ppt':'None'},
{'type':'audio','loc':'openai','name':"whisper-1",'rank':'op','max_tokens':32000,'schema':{'bos':[],'eos':''},'ppt':'None'},
]
- Add more model sources
- Screen Reading
- Console Control
- Screen Control
- Voice in/out
This project is licensed under the MIT License - see the LICENSE file for details.