Spaces:
Sleeping
Sleeping
Download vector_utils.py from Samir433/YoutubeRAG: direct link, hf CLI and curl.
- Browser
- Download file 692 Bytes
-
https://huggingface.co/spaces/Samir433/YoutubeRAG/resolve/main/vector_utils.py
- Command line
-
hf download hf://spaces/Samir433/YoutubeRAG/vector_utils.py
-
curl -L -o vector_utils.py https://huggingface.co/spaces/Samir433/YoutubeRAG/resolve/main/vector_utils.py
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}) |