From 08d7a3f8e2b991edf14a5383bede5a861b2abaeb Mon Sep 17 00:00:00 2001 From: kshitijdshah99 Date: Wed, 2 Oct 2024 00:22:32 +0530 Subject: [PATCH] Adding AI asistant code --- Pippy/Pippy_AI_assistant.py | 146 ++++++++++++++++++++++++++++++++++++ 1 file changed, 146 insertions(+) create mode 100644 Pippy/Pippy_AI_assistant.py diff --git a/Pippy/Pippy_AI_assistant.py b/Pippy/Pippy_AI_assistant.py new file mode 100644 index 0000000..69e68a7 --- /dev/null +++ b/Pippy/Pippy_AI_assistant.py @@ -0,0 +1,146 @@ +# Pippy's AI-Coding Assistant +# Uses Llama3.1 model from Ollama + +from langchain_community.vectorstores import FAISS +from langchain_community.embeddings import HuggingFaceEmbeddings, OpenAIEmbeddings +from langchain_community.document_loaders import PyMuPDFLoader, TextLoader +from langchain.chains import RetrievalQA +from langchain_core.prompts import ChatPromptTemplate +from langchain_ollama.llms import OllamaLLM +import warnings +import os + +# Suppress future warnings related to deprecations +warnings.filterwarnings("ignore", category=FutureWarning) +warnings.filterwarnings("ignore", category=DeprecationWarning) + + +# Paths to your PDF/text documents to be retrieved in the RAG model (you can add more paths to this list) +document_paths = [ + '/home/kshitij/Downloads/AI-model/Pygame Documentation.pdf', + '/home/kshitij/Downloads/AI-model/AI-model(Streamlitfree)/Python GTK+3 Documentation.pdf', +] + + +def setup_vectorstore(file_paths): + """ + Set up a vector store from the provided document files. + + Args: + file_paths (list): List of paths to document files. + + Returns: + retriever: A retriever object for document retrieval. + """ + try: + all_documents = [] + for file_path in file_paths: + if os.path.exists(file_path): + print(f"Loading document from: {file_path}") + # Load PDF or text documents based on file extension + if file_path.endswith(".pdf"): + loader = PyMuPDFLoader(file_path) + else: + loader = TextLoader(file_path) + + documents = loader.load() + print(f"Loaded {len(documents)} documents from {file_path}.") + all_documents.extend(documents) # Combine documents from all files + else: + print(f"File not found: {file_path}") + + # Create embeddings using HuggingFace or OpenAI embeddings + embeddings = HuggingFaceEmbeddings() # You can switch to OpenAIEmbeddings if needed + + # Create a vector store with all combined documents + vector_store = FAISS.from_documents(all_documents, embeddings) + + # Create a retriever + retriever = vector_store.as_retriever() + return retriever + + except Exception as e: + print(f"Failed to set up the retriever: {e}") + return None + + +# Defining a system prompt to prioritize coding-specific responses and guide the model +system_prompt = """ +You are a highly intelligent Python coding assistant with access to both general knowledge and specific Pygame documentation. +1. You only have to answer Python and GTK based coding queries. +2. Prioritize answers based on the documentation when the query is related to it. However make sure you are not biased towards documentation provided to you. +3. Make sure that you don't mention words like context or documentation stating what has been provided to you. +4. Provide step-by-step explanations wherever applicable. +5. If the documentation does not contain relevant information, use your general knowledge. +6. Always be clear, concise, and provide examples where necessary. +""" + + +# Set up the RAG model with system prompt +template = f"""{system_prompt} +Question: {{question}} +Answer: Let's think step by step. +""" +prompt = ChatPromptTemplate.from_template(template) +model = OllamaLLM(model="llama3.1") + +# Initialize the vector store retriever using multiple documents +retriever = setup_vectorstore(document_paths) + +# Used to calculate relevance of a query +def get_relevant_document(query, retriever, threshold=0.5): + """ + Check if the query is related to the document by using the retriever. + + Args: + query (str): The user query. + retriever: The document retriever. + threshold (float): The confidence threshold to decide if a query is related to the document. + + Returns: + result (dict): Retrieved document information and similarity score. + """ + results = retriever.get_relevant_documents(query) + if results and len(results) > 0: + # Check the confidence score of the first result (assuming retriever returns sorted results) + top_result = results[0] + score = top_result.metadata.get("score", 0.0) # Default to 0.0 if no score is found + + # Return the result and its score if it meets the threshold + if score >= threshold: + return top_result, score + return None, 0.0 + +if retriever: + # Combine the retriever and model into a Retrieval-Augmented Generation (RAG) chain + rag_chain = RetrievalQA.from_chain_type(llm=model, chain_type="stuff", retriever=retriever) + + print("RAG model setup completed successfully!") + + while True: + # Capture user input from the terminal + question = input("Enter your question (or type 'exit' to quit): ").strip() + + if question.lower() == 'exit': + print("Exiting the application. Goodbye!") + break + + # Check if the query is relevant to the document content + doc_result, relevance_score = get_relevant_document(question, retriever) + + if doc_result: + print(f"Document is relevant (Score: {relevance_score:.2f}). Using document-specific response...") + response = rag_chain({"query": question}) + if 'result' in response: + print(f"Document Response: {response['result']}") + else: + print("No result found in documents, trying general knowledge...") + response = model.invoke(question) + print(f"Response: {response}") + else: + print(f"Document is not relevant (Score: {relevance_score:.2f}). Using general knowledge response...") + response = model.invoke(question) + print(f"Response: {response}") + +else: + print("Unable to set up the retriever. Please check the documents and try again.") \ No newline at end of file