LoCoBot is a simple CLI chat bot that runs 100% locally. Given the directory of a software project on your machine, it can be a useful coding assistant. This is basically a LLama3.2 wrapper (although the model can be easily changed in the src/locobot.py file). This was mainly made to learn more about running contextualized LLMs locally, and to have a coding chat bot which I know for certain won't send my information to some big Tech company.
Ollama
Start by installing ollama: https://ollama.com/download
Then, pull llama3.2, and the embedding model mxbai-embed-large.
ollama pull llama3.2
ollama pull mxbai-embed-largePython virtual environment
Note: python >= 3.10 is required.
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
source .venv/bin/activateIf, while trying to run LoCoBot, you encounter some error related to sqlite3, try running:
python -m pip install pysqlite3-binaryTo run LoCoBot, simply use
python src/main.py config.jsonYou must pass a config json which specifies the directory of a software project on your machine, and the files/paths to include and exclude. See example.json for reference.
There is a set of specific prompts that LoCoBot interprets a as commands. These include commands to save the conversation to a markdown file, listing the files LoCoBot is using for context, etc. Use the prompt help to get a comprehensive list of these commands.