app.py

from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate,MessagesPlaceholder
from langchain_core.runnables import RunnablePassthrough
from langchain_huggingface.embeddings import HuggingFaceEmbeddings
from langchain.retrievers.document_compressors import EmbeddingsFilter
from langchain.retrievers import ContextualCompressionRetriever
from langchain.retrievers import EnsembleRetriever
from langchain_community.vectorstores import FAISS
from langchain_groq import ChatGroq
from langchain import hub
import pickle
import os
import gradio as gr
import spaces

GROQ_API_KEY="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"
os.environ['GROQ_API_KEY'] = GROQ_API_KEY


embed_model = HuggingFaceEmbeddings(model_name="Alibaba-NLP/gte-multilingual-base", model_kwargs={"trust_remote_code":True, "device": "cuda"})
llm = ChatGroq(
    model="llama-3.1-8b-instant",  
    temperature=0.0,
    max_tokens=1024, 
    max_retries=2
)

excel_vectorstore = FAISS.load_local(folder_path="./faiss_excel_doc_index", embeddings=embed_model, allow_dangerous_deserialization=True)
word_vectorstore = FAISS.load_local(folder_path="./faiss_recursive_split_word_doc_index", embeddings=embed_model, allow_dangerous_deserialization=True)
excel_vectorstore.merge_from(word_vectorstore)
combined_vectorstore = excel_vectorstore

with open('combined_recursive_keyword_retriever.pkl', 'rb') as f:
    combined_keyword_retriever = pickle.load(f)
    combined_keyword_retriever.k = 10

semantic_retriever = combined_vectorstore.as_retriever(search_type="mmr", search_kwargs={'k': 10, 'lambda_mult': 0.25})


# initialize the ensemble retriever
ensemble_retriever = EnsembleRetriever(
    retrievers=[combined_keyword_retriever, semantic_retriever], weights=[0.5, 0.5]
)


embeddings_filter = EmbeddingsFilter(embeddings=embed_model, similarity_threshold=0.6)
compression_retriever = ContextualCompressionRetriever(
    base_compressor=embeddings_filter, base_retriever=ensemble_retriever
)

template = """
User: You are an AI Assistant that follows instructions extremely well.
Please be truthful and give direct answers. Please tell 'I don't know' if user query is not in CONTEXT

Keep in mind, you will lose the job, if you answer out of CONTEXT questions

CONTEXT: {context}
Query: {question}

Remember only return AI answer
Assistant:
"""

prompt = ChatPromptTemplate.from_template(template)
output_parser = StrOutputParser()

def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)


rag_chain = (
    {"context": compression_retriever.with_config(run_name="Docs") | format_docs, "question": RunnablePassthrough()}
    | prompt
    | llm
    | output_parser
)





# zero = torch.Tensor([0]).cuda()

# @spaces.GPU
def get_response(question, history):
    # print(question)
    curr_ans = ""
    for chunk in rag_chain.stream(question):
        curr_ans += chunk
        yield curr_ans

example_questions = [
    "الموسم المناسب لزراعة الذرة العلفية ؟",
    "ما هي الاحتياجات المائية لتربية الحيوانات؟",
    "ما هي خطوات إنتاج الشتلات؟",
    "الموسم المناسب لزراعة الطماطم في الحقل المكشوف بدولة الإمارات؟",
    "شروط اختيار مكان منحل العسل؟",
    "ما هو تقييم مطعم قصر نجد؟",
    "ما كمية أعلاف الجت المستلمة في منطقة الظفرة عام 2022",
    "ما مساحات المزارع المروية بالتنقيط في منطقة الرحبة عام 2020",
    "في إمارة أبوظبي في عام 2022، هل نسبة العينات الغذائية الغير مطابقة من إجمالي العينات أعلى في العينات المحلية أم العينات المستوردة"
]
with gr.Blocks() as demo:
    gr.Markdown(
        """
        # ADAFSA RAG Chatbot Demo
        """
    )
    chatbot = gr.Chatbot(placeholder="<strong>ADAFSA-RAG Chatbot</strong>")
    gr.ChatInterface(
        title="",
        fn=get_response, 
        chatbot=chatbot,
        examples=example_questions,
    )
    
demo.launch()