Build the Wall-E Robot with the cutting edge AI tools.
And you can talk with Robot, like "What can you see on your right?", and then it turn right, and speak out what she can see.
- OpenAI
- Cohere AI
- Langchain
- Edge-TTS
- Vision Capture(Salesforce BLIP)
- Idea from: https://github.com/xISSAx/Alpha-Co-Vision
- Optimized: https://github.com/pjq/Alpha-Co-Vision
- Google Search
- Weather
- Wolfram
- Whisper
And generally speaking, it can integrate all the tools build by Langchain.
git clone git@github.com:pjq/wall-e-robot.git
pip install -r requirements.txt
{
"cohere_api_key": "",
"openai_api_key": "",
"enable_openai": true,
"openai_api_base_url": "",
"enable_mps": true,
"blip_model": "Salesforce/blip-image-captioning-base",
"whisper_cpp_path":"",
"edge_tts_enable": true,
"car_api_url": "",
"edge_tts_voice": "en-US-JennyNeural",
}
python wall-e.py --config myconfig.json
Several years ago, I build the smart car with Raspberry Pi
And it supports call the Restful API to control the car which can be used by the LLM.
And basically, it need to support the following params.
class CarAction:
SUPPORTED_ACTIONS = ("stop", "left", "right", "up", "down")
def __init__(self):
self.action = ""
self.duration = 0
self.speed = 0
self.angle = 0
- https://github.com/pjq/rpi
- https://github.com/xISSAx/Alpha-Co-Vision
- https://github.com/pjq/Alpha-Co-Vision
ImportError: cannot import name 'BlipProcessor' from 'transformers' (/Users/i329817/miniconda3/lib/python3.10/site-packages/transformers/__init__.py)
You can install it
pip install git+https://github.com/huggingface/transformers
conda create --name wall-e python=3.10
conda activate wall-e
pip install -r requirements.txt
conda deactivate wall-e