LLM Augmentation

Use your own data with the power of an LLM

RAG (Retrieval Augmented Generation)

Way to add data to an llm model without retraining it. needed when llm model has a cutoff date or for using sensitive data.

Vector Embeddings

A format of data which holds semantic information.

Semantic Search

Way to identiify things which are similar in meaning, not in the characters. eg: dog is similar to cat; dog is not similar to dot.

Useful commands for setup

ollama --version
ollama list
ollama pull llama3.2:1b

poetry init --no-interaction
poetry add faiss-cpu langchain langchain-community langchain_experimental sentence-transformers langchain-ollama langchain-huggingface fastapi uvicorn python-multipart

To use deepseek r1 ollama pull deepseek-r1:1.5b

running the files

python script.py

or

uvicorn server:app --reload