Try our beta now! ( currently out of order :-( )
If you're connecting to a local LLM with ollama, you may also need to run ollama serve in the imagination server (or wherever ollama is hosted) to launch the instance.
python3 -m venv reality-env
source reality-env/bin/activate
pip install -r requirements.txt
python main.py --testIf you're running on a server where ollama is running, make the LOCAL_PORT in the .env 11434 and leave the imagination (remote) server variables blank. Then run python main.py --test --no-tunnel.
IMAGINATION_IP=
IMAGINATION_PORT=11434 # ollama port
LOCAL_PORT= # arbitrary unless ollama port when running the program in a server where ollama is running.
SSH_USERNAME=
SSH_KEYFILE=
If you're running on a cuttlefish, run this:
python main.py --cuttlefish --no-tunnel --model openai
Read more here
I would have added frontend functionality, asynchronous programming and websockets from the get-go. I would have grouped prompts more from the beginning based on the associated step. I also would have started using Pydantic types sooner. I also shouldn't have dismissed the benefits of using a tool like Langchain but I'm happy to have had the experience to write a lot of stuff myself.
Also I'm not using Poetry again, I forgot about good old venv in my time away from Python.