End-to-end large language model with retrievement augmentation, for free use.
The RAG components come from Bible repositories:
- Many public domain translations of Bibles: https://github.com/gratis-bible/bible
- Interlinear Bibles: https://github.com/ivandustin/bible
The exploratory plan is presented in a jupyter notebook format for easy presentation.
- Cookie Cutter
- Start Learning
- DS MLE
- DS MLE saved to .csv
This model is first deployed to a local LLM. This requires LLM studio to be running with a viable endpoint, or an equivalent API.
This project is meant to be viewable and consumable by anyone without the need for a commercial LLM subscription. To Do:
- README refinement
- GitHub optimaization
- inline visualizations
- Install Ollama from this website: https://ollama.com/download/windows
- Open Command Prompt and type
olamma pull llama3to install the Llama3 model (or install minstral). - confirm that Ollama is running by checking localhost:11434, which should say, "Ollama is running". To Do:
- vector database
- fine tuning
- RAG setup
Run the following code after running pip install langchain-community in your virtual environment:
from langchain_community.llms import Ollama
llm = Ollama(model="llama3")
llm.invoke("Write a haiku.")
I like to test it by writing a haiku because I know the output will be quick.
If this code works, then you have correctly setup Ollama on your local machine.
This project is meant to provide easy reference to Bible content and compare translations and languages.
An LLM only understands language as a vector of integers representing word-tokens. Therefore, the human-readable text corpus must be translated into vectors. For user convenience, these vectors are stored in a vector database as a flat file.
It's finally time to enjoy the fruits of your labour! Your LLM is able to retrieve information from your text corpus.