File size: 3,738 Bytes
72aebc1 b676165 72aebc1 b676165 1393d5f 46f3018 f6c70f7 46f3018 f6c70f7 b676165 72aebc1 b676165 72aebc1 b676165 72aebc1 b676165 72aebc1 b676165 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 |
import os
import gradio as gr
from langchain.vectorstores import Chroma
from transformers import RagTokenizer, RagSequenceForGeneration
from sentence_transformers import SentenceTransformer
from langchain.chains.question_answering import load_qa_chain
from langchain.llms import HuggingFaceLLM
#Konstanten
ANTI_BOT_PW = os.getenv("CORRECT_VALIDATE")
# Setzen des Hugging Face Tokens als Umgebungsvariable
os.environ["HUGGINGFACEHUB_API_TOKEN"] = os.getenv("HF_READ")
# Initialisierung des Sentence-BERT Modells für die Embeddings
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
# Initialisierung von Tokenizer und RAG Modell
tokenizer = RagTokenizer.from_pretrained("facebook/rag-sequence-nq", use_auth_token=True)
model = RagSequenceForGeneration.from_pretrained("facebook/rag-sequence-nq", use_auth_token=True)
# Verbindung zur Chroma DB und Laden der Dokumente
chroma_db = Chroma(embedding_model=embedding_model, persist_directory = PATH_WORK + CHROMA_DIR)
# Erstellung eines HuggingFaceLLM Modells
llm = HuggingFaceLLM(model=model, tokenizer=tokenizer)
# Erstellen eines eigenen Retrievers mit Chroma DB und Embeddings
#retriever = chroma_db.as_retriever()
# Erstellung der RAG-Kette mit dem benutzerdefinierten Retriever
#rag_chain = RagChain(model=model, retriever=retriever, tokenizer=tokenizer, vectorstore=chroma_db)
#############################################
def document_retrieval_chroma2():
#HF embeddings -----------------------------------
#Alternative Embedding - für Vektorstore, um Ähnlichkeitsvektoren zu erzeugen - die ...InstructEmbedding ist sehr rechenaufwendig
embeddings = HuggingFaceInstructEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2", model_kwargs={"device": "cpu"})
#etwas weniger rechenaufwendig:
#embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2", model_kwargs={"device": "cpu"}, encode_kwargs={'normalize_embeddings': False})
#oder einfach ohne Langchain:
#embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
#ChromaDb um die embedings zu speichern
db = Chroma(embedding_function = embeddings, persist_directory = PATH_WORK + CHROMA_DIR)
print ("Chroma DB bereit ...................")
return db
def get_rag_response(question):
# Abfrage der relevanten Dokumente aus Chroma DB
docs = chroma_db.search(question, top_k=5)
passages = [doc['text'] for doc in docs]
links = [doc.get('url', 'No URL available') for doc in docs]
# Generieren der Antwort
answer = llm(question, docs)
# Zusammenstellen der Ausgabe
response = {
"answer": answer,
"documents": [{"link": link, "passage": passage} for link, passage in zip(links, passages)]
}
return response
def chatbot_response (user_input, chat_history=[]):
response = get_rag_response(user_input)
answer = response['answer']
documents = response['documents']
doc_links = "\n\n".join([f"Link: {doc['link']} \nAuszüge der Dokumente: {doc['passage']}" for doc in documents])
bot_response = f"{answer} \n\nRelevante Dokumente: \n{doc_links}"
chat_history.append((user_inptu, bot_response))
return chat_history, chat_history
#############################
#GUI.........
def user (user_input, history):
return "", history + [[user_input, None]]
with gr.Blocks() as chatbot:
chat_interface = gr.Chatbot()
msg = gr.Textbox()
clear = gr.Button("Löschen")
#Buttons listener
msg.submit(user, [msg, chat_interface], [msg, chat_interface], queue = False). then(chatbot_response, [msg, chat_interface], [chat_interface, chat_interface])
clear.click(lambda: None, None, chat_interface, queue=False)
chatbot.launch() |