import os
from langchain_community.document_loaders import DirectoryLoader
from langchain_community.vectorstores import FAISS
from langchain_huggingface import HuggingFaceEmbeddings
from rag_pipeline import process_single_file # Re-use our processing logic
from config import settings

def main():
    print("Starting creation of the fallback FAISS index...")

    # This assumes your general documents are in a folder named 'fallback_documents'
    # You can change this path as needed.
    source_directory = "fallback_documents"
    all_docs = []

    for filename in os.listdir(source_directory):
        file_path = os.path.join(source_directory, filename)
        if os.path.isfile(file_path):
            with open(file_path, "rb") as f:
                content = f.read()
                docs = process_single_file(content, filename)
                all_docs.extend(docs)

    if not all_docs:
        print("No documents found to index. Exiting.")
        return

    print(f"Generating embeddings for {len(all_docs)} chunks in the fallback documents...")
    embeddings = HuggingFaceEmbeddings(model_name=settings.EMBEDDING_MODEL_NAME)

    # Create the FAISS index from the documents
    vector_store = FAISS.from_documents(all_docs, embeddings)

    # Save the index to the path specified in your config
    # Ensure this path is inside your persistent storage volume (e.g., /data)
    fallback_path = settings.FALLBACK_INDEX_PATH
    os.makedirs(fallback_path, exist_ok=True)
    vector_store.save_local(fallback_path)

    print(f"Fallback index created and saved successfully at: {fallback_path}")

if __name__ == "__main__":
    main()