rag.py

from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_pinecone import PineconeVectorStore
from langchain_core.documents import Document

from langchain_openai import ChatOpenAI
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import PromptTemplate
from uuid import uuid4

import unicodedata

def remove_non_standard_ascii(input_string: str) -> str:
    normalized_string = unicodedata.normalize('NFKD', input_string)
    return ''.join(char for char in normalized_string if 'a' <= char <= 'z' or 'A' <= char <= 'Z' or char.isdigit() or char in ' .,!?')




def get_text_from_content_for_doc(content):
    text = ""
    for page in content:
        text += content[page]["texte"]
    return text

def get_text_from_content_for_audio(content):
    return content["transcription"]


def get_text_chunks(text):
    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=500, # the character length of the chunck
        chunk_overlap=100, # the character length of the overlap between chuncks
        length_function=len # the length function - in this case, character length (aka the python len() fn.)
    )
    chunks = text_splitter.split_text(text)
    return chunks

def get_vectorstore(text_chunks,filename, file_type,namespace,index,enterprise_name):
    try:
        embedding = OpenAIEmbeddings(model="text-embedding-3-large")
        vector_store = PineconeVectorStore(index=index, embedding=embedding,namespace=namespace)

        file_name = filename.split(".")[0].replace(" ","_").replace("-","_").replace(".","_").replace("/","_").replace("\\","_").strip()

        documents = []
        uuids = []

        for i, chunk in enumerate(text_chunks):
            clean_filename = remove_non_standard_ascii(file_name)

            document = Document(
            page_content=chunk,
            metadata={"filename":filename,"file_type":file_type, "filename_id":clean_filename, "entreprise_name":enterprise_name},
            )
    
            uuid = f"{clean_filename}_{i}"
            uuids.append(uuid)
            documents.append(document)
        
        vector_store.add_documents(documents=documents, ids=uuids)

        return {"filename_id":clean_filename}
    
    except Exception as e:
        print(e)
        return False
    
def get_retreive_answer(enterprise_id,prompt,index,common_id):
    try:
        
        embedding = OpenAIEmbeddings(model="text-embedding-3-large")
        vector_store = PineconeVectorStore(index=index, embedding=embedding,namespace=enterprise_id)

        retriever = vector_store.as_retriever(
            search_type="similarity_score_threshold",
            search_kwargs={"k": 3, "score_threshold": 0.6},
        )

        if common_id:
            vector_store_commun = PineconeVectorStore(index=index, embedding=embedding,namespace=common_id)

            retriever_commun = vector_store_commun.as_retriever(
                search_type="similarity_score_threshold",
                search_kwargs={"k": 3, "score_threshold": 0.1},
            )

            response = retriever.invoke(prompt) + retriever_commun.invoke(prompt)

        else:
            response = retriever.invoke(prompt)

        return response
    
    except Exception as e:
        print(e)
        return False
    
    
def generate_response_via_langchain(query: str, stream: bool = False, model: str = "gpt-4o",context:str="",messages = [],style:str="formel",tonality:str="neutre",template:str = ""):
    # Define the prompt template
    if template == "":
        template = "En tant qu'IA experte en marketing, réponds avec un style {style} et une tonalité {tonality} dans ta communcation, sachant le context suivant: {context}, et l'historique de la conversation, {messages}, {query}"


    prompt = PromptTemplate.from_template(template)
    
    # Initialize the OpenAI LLM with the specified model
    llm = ChatOpenAI(model=model,temperature=0)
    
    # Create an LLM chain with the prompt and the LLM
    llm_chain = prompt | llm | StrOutputParser()

    if stream:
        # Return a generator that yields streamed responses
        return llm_chain.astream({ "query": query, "context": context, "messages": messages, "style": style, "tonality": tonality })
    
    # Invoke the LLM chain and return the result
    return llm_chain.invoke({"query": query, "context": context, "messages": messages, "style": style, "tonality": tonality})



def setup_rag(file_type,content):
    if file_type == "pdf":
        text = get_text_from_content_for_doc(content)
    elif file_type == "audio":
        text = get_text_from_content_for_audio(content)
    

    chunks = get_text_chunks(text)

    vectorstore = get_vectorstore(chunks)
    
    return vectorstore


def prompt_reformatting(prompt:str,context,query:str,style="formel",tonality="neutre"):
    if context == "":
        return prompt.format(context="Pas de contexte pertinent",messages="",query=query,style=style,tonality=tonality)
    
    docs_names = []
    for chunk in context:
        print(chunk.metadata)
        chunk_name = chunk.metadata["filename"]
        if chunk_name not in docs_names:
            docs_names.append(chunk_name)
    context = ", ".join(docs_names)

    prompt = prompt.format(context=context,messages="",query=query,style=style,tonality=tonality)
    return prompt