app.py

import gradio as gr
import os
from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_huggingface import HuggingFaceEmbeddings, HuggingFaceEndpoint
from langchain_community.vectorstores import FAISS
from langchain.chains import ConversationalRetrievalChain
from langchain.memory import ConversationBufferMemory

# API-Token
api_token = os.getenv("HF_TOKEN")

# LLM-Optionen
list_llm = ["google/flan-t5-small", "google/flan-t5-base"]

# Dokumente laden und aufteilen
def load_doc(list_file_path):
    if not list_file_path:
        return [], "Fehler: Keine Dokumente gefunden!"
    loaders = [PyPDFLoader(x) for x in list_file_path]
    documents = []
    for loader in loaders:
        documents.extend(loader.load())
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1024, chunk_overlap=64)
    return text_splitter.split_documents(documents)

# Vektor-Datenbank erstellen
def create_db(splits):
    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
    return FAISS.from_documents(splits, embeddings)

# Datenbank initialisieren
def initialize_database(list_file_obj):
    if not list_file_obj:
        return None, "Fehler: Keine Dateien hochgeladen!"
    list_file_path = [x.name for x in list_file_obj if x is not None]
    doc_splits = load_doc(list_file_path)
    vector_db = create_db(doc_splits)
    return vector_db, "Datenbank erfolgreich erstellt!"

# LLM-Kette initialisieren
def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
    if vector_db is None:
        return None, "Fehler: Keine Vektordatenbank verfügbar."
    llm = HuggingFaceEndpoint(
        repo_id=llm_model,
        huggingfacehub_api_token=api_token,
        temperature=temperature,
        max_new_tokens=max_tokens,
        top_k=top_k,
    )
    memory = ConversationBufferMemory(memory_key="chat_history", output_key="answer", return_messages=True)
    retriever = vector_db.as_retriever()
    return ConversationalRetrievalChain.from_llm(
        llm, retriever=retriever, chain_type="stuff", memory=memory, return_source_documents=True
    )

# LLM initialisieren
def initialize_LLM(llm_option, llm_temperature, max_tokens, top_k, vector_db):
    if vector_db is None:
        return None, "Datenbank wurde nicht erstellt!"
    llm_name = list_llm[llm_option]
    qa_chain = initialize_llmchain(llm_name, llm_temperature, max_tokens, top_k, vector_db)
    return qa_chain, "QA-Kette initialisiert. Chatbot ist bereit!"

# Konversation
def conversation(qa_chain, message, history):
    if qa_chain is None:
        return None, [{"role": "system", "content": "Die QA-Kette wurde nicht initialisiert."}], history
    if not message.strip():
        return qa_chain, [{"role": "system", "content": "Bitte eine Frage eingeben!"}], history
    response = qa_chain.invoke({"question": message, "chat_history": history})
    response_text = response.get("answer", "Keine Antwort verfügbar.")
    sources = [doc.metadata["source"] for doc in response.get("source_documents", [])]
    formatted_response = history + [{"role": "assistant", "content": response_text}]
    return qa_chain, formatted_response, formatted_response

# Demo erstellen
def demo():
    with gr.Blocks() as demo:
        vector_db = gr.State()
        qa_chain = gr.State()
        gr.Markdown("<center><h1>PDF-Chatbot mit kostenlosen Modellen</h1></center>")
        document = gr.Files(label="PDF-Dokument hochladen")
        db_btn = gr.Button("Erstelle Vektordatenbank")
        llm_btn = gr.Radio(["Flan-T5 Small", "Flan-T5 Base"], label="Verfügbare LLMs", value="Flan-T5 Small", type="index")
        slider_temperature = gr.Slider(0.01, 1.0, 0.5, label="Temperature")
        slider_maxtokens = gr.Slider(128, 2048, 512, label="Max Tokens")
        slider_topk = gr.Slider(1, 10, 3, label="Top-k")
        qachain_btn = gr.Button("Initialisiere QA-Chatbot")
        chatbot = gr.Chatbot(label="Chatbot", height=400)
        msg = gr.Textbox(label="Frage stellen")
        submit_btn = gr.Button("Absenden")

        db_btn.click(initialize_database, inputs=[document], outputs=[vector_db])
        qachain_btn.click(initialize_LLM, inputs=[llm_btn, slider_temperature, slider_maxtokens, slider_topk, vector_db], outputs=[qa_chain])
        submit_btn.click(conversation, inputs=[qa_chain, msg, chatbot], outputs=[qa_chain, chatbot, chatbot])

    demo.launch(debug=True)

if __name__ == "__main__":
    demo()