Download main.py from Manav0603/project2: direct link, hf CLI and curl.
- Browser
- Download file 3.44 kB
-
https://huggingface.co/spaces/Manav0603/project2/resolve/main/main.py
- Command line
-
hf download hf://spaces/Manav0603/project2/main.py
-
curl -L -o main.py https://huggingface.co/spaces/Manav0603/project2/resolve/main/main.py
3.44 kB
| import os | |
| import shutil | |
| from langchain_community.document_loaders import PyPDFLoader | |
| from langchain_community.embeddings import HuggingFaceEmbeddings | |
| from langchain_community.vectorstores import Chroma | |
| from langchain.text_splitter import RecursiveCharacterTextSplitter | |
| from langchain.chains import RetrievalQA | |
| from langchain_community.llms import HuggingFaceEndpoint | |
| # ---------------- Global QA object ---------------- | |
| qa = None | |
| # ---------------- Initialize LLM ---------------- | |
| def initialize_llm(): | |
| """ | |
| Initialize HuggingFaceEndpoint LLM for Mistral-7B | |
| β οΈ Replace 'hf_your_actual_token_here' with your actual Hugging Face token | |
| """ | |
| hf_token = "token2" # <-- Your Hugging Face token | |
| llm = HuggingFaceEndpoint( | |
| endpoint_url="https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2", | |
| huggingfacehub_api_token=hf_token, | |
| task="text2text-generation", | |
| model_kwargs={"temperature": 0.1, "max_length": 512} | |
| ) | |
| print("β Mistral-7B LLM initialized") | |
| return llm | |
| # Initialize global LLM | |
| llm = initialize_llm() | |
| # ---------------- PDF Processing ---------------- | |
| def process_pdf(pdf_file_path): | |
| """Load PDF, create embeddings, and initialize QA system""" | |
| global qa | |
| if llm is None: | |
| return "β Error: Failed to initialize LLM. Check your HF token." | |
| try: | |
| # Load PDF | |
| loader = PyPDFLoader(pdf_file_path) | |
| documents = loader.load() | |
| if not documents: | |
| return "β Error: No content found in PDF" | |
| # Split text into chunks | |
| text_splitter = RecursiveCharacterTextSplitter( | |
| chunk_size=1000, | |
| chunk_overlap=200, | |
| length_function=len | |
| ) | |
| docs = text_splitter.split_documents(documents) | |
| docs = [doc for doc in docs if doc.page_content.strip()] | |
| if not docs: | |
| return "β Error: PDF has no extractable text chunks" | |
| # Create embeddings | |
| embeddings = HuggingFaceEmbeddings( | |
| model_name="sentence-transformers/all-MiniLM-L6-v2", | |
| model_kwargs={"device": "cpu"} | |
| ) | |
| # Clear previous Chroma cache | |
| if os.path.exists("./chroma_cache"): | |
| shutil.rmtree("./chroma_cache") | |
| # Create vectorstore | |
| vectorstore = Chroma.from_documents( | |
| docs, embeddings, persist_directory="./chroma_cache" | |
| ) | |
| # Initialize QA system | |
| qa = RetrievalQA.from_chain_type( | |
| llm=llm, # β Pass HuggingFaceEndpoint instance here | |
| chain_type="stuff", | |
| retriever=vectorstore.as_retriever(search_kwargs={"k": 3}), | |
| return_source_documents=False | |
| ) | |
| return f"β PDF processed successfully! {len(docs)} text chunks ready." | |
| except Exception as e: | |
| return f"β Error processing PDF: {str(e)}" | |
| # ---------------- Query PDF ---------------- | |
| def query_pdf(question): | |
| """Query the processed PDF""" | |
| global qa | |
| if qa is None: | |
| return "β No PDF processed yet. Please upload and process a PDF first." | |
| if not question.strip(): | |
| return "β οΈ Please ask a valid question." | |
| try: | |
| result = qa.invoke({"query": question}) | |
| return result.get("result", "No answer found") | |
| except Exception as e: | |
| return f"β Error answering question: {str(e)}" |