YoutubeRAG / vector_utils.py
Samir433's picture
Upload 6 files
dc759f5 verified
Raw History Blame Contribute Delete
692 Bytes
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
splitter = RecursiveCharacterTextSplitter(chunk_size=900, chunk_overlap=100)
def build_vector_store(transcript: str, cache_folder: str, model_name: str):
# cache_folder is ignored for Hugging Face Spaces
docs = splitter.create_documents([transcript])
embeddings = HuggingFaceEmbeddings(model_name=model_name)
return FAISS.from_documents(docs, embeddings)
def get_retriever(vector_store, k: int = 4):
return vector_store.as_retriever(search_type="similarity", search_kwargs={"k": k})