import gradio as gr import os import fitz # PyMuPDF import tempfile from langchain_community.embeddings import HuggingFaceEmbeddings from langchain_community.vectorstores import Chroma from langchain.text_splitter import RecursiveCharacterTextSplitter from langchain_community.document_loaders import TextLoader from langchain.chains import RetrievalQA from langchain.llms.base import LLM from typing import List, Optional from groq import Groq # --------------------- Custom LLM Class --------------------- class GroqLLM(LLM): model: str = "llama3-8b-8192" api_key: str = os.getenv("GROQ_API_KEY") # Get securely from environment temperature: float = 0.0 def _call(self, prompt: str, stop: Optional[List[str]] = None) -> str: client = Groq(api_key=self.api_key) messages = [ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": prompt} ] response = client.chat.completions.create( model=self.model, messages=messages, temperature=self.temperature, ) return response.choices[0].message.content @property def _llm_type(self) -> str: return "groq-llm" # --------------------- Global State --------------------- qa_chain = None summary_text = "" # --------------------- PDF Processor --------------------- def process_pdf(file): global qa_chain, summary_text # Save uploaded file temp_dir = tempfile.mkdtemp() pdf_path = os.path.join(temp_dir, "uploaded.pdf") with open(pdf_path, "wb") as f: f.write(file.read()) # Extract text using PyMuPDF doc = fitz.open(pdf_path) text = "\n".join([page.get_text() for page in doc]) # Save as .txt for loader txt_path = os.path.join(temp_dir, "temp.txt") with open(txt_path, "w", encoding="utf-8") as f: f.write(text) # LangChain pipeline loader = TextLoader(txt_path) documents = loader.load() text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50) docs = text_splitter.split_documents(documents) embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2") vectorstore = Chroma.from_documents(docs, embeddings, persist_directory=os.path.join(temp_dir, "chroma_db")) retriever = vectorstore.as_retriever() groq_llm = GroqLLM() qa_chain = RetrievalQA.from_chain_type( llm=groq_llm, retriever=retriever, return_source_documents=True ) # Generate auto-summary summary_prompt = "Summarize what this document is about in 5-6 lines." summary_text = qa_chain({"query": summary_prompt})["result"] return summary_text # --------------------- Question Answering --------------------- def answer_question(query): global qa_chain if not qa_chain: return "Please upload and process a PDF first." result = qa_chain({"query": query}) return result["result"] # --------------------- Gradio UI --------------------- with gr.Blocks() as demo: gr.Markdown("## 📄 PDF Summarizer & Q&A using Groq LLM (LLaMA3-8B)") with gr.Row(): pdf_input = gr.File(label="Upload a PDF", type="binary") summary_output = gr.Textbox(label="Auto Summary", lines=8) with gr.Row(): process_button = gr.Button("Process PDF") with gr.Row(): question = gr.Textbox(label="Ask a question about the PDF") answer = gr.Textbox(label="Answer", lines=6) process_button.click(fn=process_pdf, inputs=pdf_input, outputs=summary_output) question.submit(fn=answer_question, inputs=question, outputs=answer) demo.queue().launch()