detect_intent.py

from langchain_ollama import OllamaEmbeddings, ChatOllama
from langchain_chroma import Chroma
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnablePassthrough
from langchain_core.output_parsers import StrOutputParser

def format_docs(docs):
    return "\n\n".join(f"Sentence: {doc.page_content}\nIntent: {doc.metadata.get('intent', 'Unknown')}" for doc in docs)

def detect_intent_with_context(sentence):
    messages = []
    persist_directory = "db-intents"
    local_embeddings = OllamaEmbeddings(model="llama3.1:8b")
    
    vectorstore = Chroma(persist_directory=persist_directory, embedding_function=local_embeddings)
    
    docs = vectorstore.similarity_search(sentence)
    if not docs:
        messages.append("No relevant intent was found.")
        return {"response": "No matching intent found.", "messages": messages}
    
    # Define the RAG prompt template for intent detection
    INTENT_TEMPLATE = """
    You are an assistant for detecting the intent of a sentence. Use the following pieces of retrieved context to determine the intent. If you cannot determine the intent, just say that you don't know. 

    <context>
    {context}
    </context>

    What is the intent of the following sentence? Just reply with the intent name and nothing else as response.

    {sentence}"""
    
    intent_prompt = ChatPromptTemplate.from_template(INTENT_TEMPLATE)
    model = ChatOllama(model="llama3.1:8b")
    
    chain = (
        RunnablePassthrough.assign(context=lambda input: format_docs(input["context"]))
        | intent_prompt
        | model
        | StrOutputParser()
    )
    
    response = chain.invoke({"context": docs, "sentence": sentence})
    return response