app.py

from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
from langchain_community.document_loaders import PyPDFLoader
import gradio as gr

def warn(*args, **kwargs):
    pass
import warnings
warnings.warn = warn
warnings.filterwarnings('ignore')

## LLM
def get_llm():
    model_name = "TheBloke/guanaco-7B-GPTQ"
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", trust_remote_code=True)
    text_generator = pipeline(
        "text-generation",
        model=model,
        tokenizer=tokenizer,
        max_new_tokens=256,
        temperature=0.5,
        do_sample=True,
    )
    return text_generator

## Document loader
def document_loader(file):
    loader = PyPDFLoader(file)
    loaded_document = loader.load()
    full_text = "\n".join([doc.page_content for doc in loaded_document])
    return full_text

## QA function without RAG
def simple_qa(file, query):
    llm = get_llm()
    document_text = document_loader(file)
    prompt = f"Document:\n{document_text}\n\nQuestion: {query}\nAnswer:"
    response = llm(prompt)
    return response[0]['generated_text']

# Create Gradio interface
simple_application = gr.Interface(
    fn=simple_qa,
    allow_flagging="never",
    inputs=[
        gr.File(label="Upload PDF File", file_count="single", file_types=['.pdf'], type="filepath"),
        gr.Textbox(label="Input Query", lines=2, placeholder="Type your question here...")
    ],
    outputs=gr.Textbox(label="Output"),
    title="Basia's simple Q&A bot (no RAG)!",
    description="Upload a PDF document and ask any question. The chatbot will try to answer using the full document without retrieval."
)

# Launch the app
simple_application.launch(server_name="127.0.0.1", server_port=7860)