{ "cells": [ { "cell_type": "markdown", "id": "c25c6e94-f3de-4367-b2bf-269ba7160977", "metadata": {}, "source": [ "## An Expert Knowledge Worker Question-Answering Agent using RAG" ] }, { "cell_type": "markdown", "id": "15169580-cf11-4dee-8ec7-3a4ef59b19ee", "metadata": {}, "source": [ "Aims\n", "- Reads README.md files and loads data using TextLoader\n", "- Splits into chunks using CharacterTextSplitter\n", "- Converts chunks into vector embeddings and creates a datastore\n", "- 2D and 3D visualisations\n", "- Langchain to set up a conversation retrieval chain" ] }, { "cell_type": "code", "execution_count": null, "id": "051cf881-357d-406b-8eae-1610651e40f1", "metadata": {}, "outputs": [], "source": [ "# imports\n", "\n", "import os\n", "import glob\n", "from dotenv import load_dotenv\n", "import gradio as gr" ] }, { "cell_type": "code", "execution_count": null, "id": "ccfd403a-5bdb-4a8c-b3fd-d47ae79e43f7", "metadata": {}, "outputs": [], "source": [ "# imports for langchain, plotly and Chroma\n", "\n", "from langchain.document_loaders import DirectoryLoader, TextLoader\n", "from langchain.text_splitter import CharacterTextSplitter\n", "from langchain.schema import Document\n", "from langchain_openai import OpenAIEmbeddings, ChatOpenAI\n", "from langchain.embeddings import HuggingFaceEmbeddings\n", "from langchain_chroma import Chroma\n", "from langchain.memory import ConversationBufferMemory\n", "from langchain.chains import ConversationalRetrievalChain\n", "import numpy as np\n", "from sklearn.manifold import TSNE\n", "import plotly.graph_objects as go\n", "import plotly.express as px\n", "import matplotlib.pyplot as plt" ] }, { "cell_type": "code", "execution_count": null, "id": "2d853868-d2f6-43e1-b27c-b8e91d06b724", "metadata": {}, "outputs": [], "source": [ "MODEL = \"gpt-4o-mini\"\n", "db_name = \"vector_db\"" ] }, { "cell_type": "code", "execution_count": null, "id": "f152fc3b-0bf4-4d51-948f-95da1ebc030a", "metadata": {}, "outputs": [], "source": [ "# Load environment variables in a file called .env\n", "\n", "load_dotenv(override=True)\n", "os.environ['OPENAI_API_KEY'] = os.getenv('OPENAI_API_KEY')" ] }, { "cell_type": "code", "execution_count": null, "id": "24e621ac-df06-4af6-a60d-a9ed7adb884a", "metadata": {}, "outputs": [], "source": [ "# Read in documents using LangChain's loaders\n", "\n", "folder = \"my-knowledge-base/\"\n", "text_loader_kwargs={'autodetect_encoding': True}\n", "\n", "loader = DirectoryLoader(folder, glob=\"**/*.md\", loader_cls=TextLoader, loader_kwargs=text_loader_kwargs)\n", "folder_docs = loader.load()\n", "\n", "for doc in folder_docs:\n", " filename_md = os.path.basename(doc.metadata[\"source\"]) \n", " filename, _ = os.path.splitext(filename_md) \n", " doc.metadata[\"filename\"] = filename\n", "\n", "documents = folder_docs \n", "\n", "text_splitter = CharacterTextSplitter(chunk_size=400, chunk_overlap=200)\n", "chunks = text_splitter.split_documents(documents)\n", "\n", "print(f\"Total number of chunks: {len(chunks)}\")\n", "print(f\"Files found: {set(doc.metadata['filename'] for doc in documents)}\")" ] }, { "cell_type": "code", "execution_count": null, "id": "f02f08ee-5ade-4f79-a500-045a8f1a532f", "metadata": {}, "outputs": [], "source": [ "# Put the chunks of data into a Vector Store that associates a Vector Embedding with each chunk\n", "\n", "embeddings = HuggingFaceEmbeddings(model_name=\"sentence-transformers/all-MiniLM-L6-v2\")\n", "\n", "# Delete if already exists\n", "\n", "if os.path.exists(db_name):\n", " Chroma(persist_directory=db_name, embedding_function=embeddings).delete_collection()\n", "\n", "# Create vectorstore\n", "\n", "vectorstore = Chroma.from_documents(documents=chunks, embedding=embeddings, persist_directory=db_name)\n", "print(f\"Vectorstore created with {vectorstore._collection.count()} documents\")" ] }, { "cell_type": "code", "execution_count": null, "id": "7f665f4d-ccb1-43fb-b901-040117925732", "metadata": {}, "outputs": [], "source": [ "# Let's investigate the vectors\n", "\n", "collection = vectorstore._collection\n", "count = collection.count()\n", "\n", "sample_embedding = collection.get(limit=1, include=[\"embeddings\"])[\"embeddings\"][0]\n", "dimensions = len(sample_embedding)\n", "print(f\"There are {count:,} vectors with {dimensions:,} dimensions in the vector store\")" ] }, { "cell_type": "code", "execution_count": null, "id": "6208a971-e8b7-48bc-be7a-6dcb82967fd2", "metadata": {}, "outputs": [], "source": [ "# pre work\n", "\n", "result = collection.get(include=['embeddings','documents','metadatas'])\n", "vectors = np.array(result['embeddings']) \n", "documents = result['documents']\n", "metadatas = result['metadatas']\n", "filenames = [metadata['filename'] for metadata in metadatas]" ] }, { "cell_type": "code", "execution_count": null, "id": "eb27bc8a-453b-4b19-84b4-dc495bb0e544", "metadata": {}, "outputs": [], "source": [ "import random\n", "def random_color():\n", " return f\"rgb({random.randint(0,255)},{random.randint(0,255)},{random.randint(0,255)})\"" ] }, { "cell_type": "code", "execution_count": null, "id": "78db67e5-ef10-4581-b8ac-3e0281ceba45", "metadata": {}, "outputs": [], "source": [ "def show_embeddings_2d(result):\n", " vectors = np.array(result['embeddings']) \n", " documents = result['documents']\n", " metadatas = result['metadatas']\n", " filenames = [metadata['filename'] for metadata in metadatas]\n", " filenames_unique = sorted(set(filenames))\n", "\n", " # color assignment\n", " color_map = {name: random_color() for name in filenames_unique}\n", " colors = [color_map[name] for name in filenames]\n", "\n", " tsne = TSNE(n_components=2, random_state=42,perplexity=4)\n", " reduced_vectors = tsne.fit_transform(vectors)\n", "\n", " # Create the 2D scatter plot\n", " fig = go.Figure(data=[go.Scatter(\n", " x=reduced_vectors[:, 0],\n", " y=reduced_vectors[:, 1],\n", " mode='markers',\n", " marker=dict(size=5,color=colors, opacity=0.8),\n", " text=[f\"Type: {t}
Text: {d[:100]}...\" for t, d in zip(filenames, documents)],\n", " hoverinfo='text'\n", " )])\n", "\n", " fig.update_layout(\n", " title='2D Chroma Vector Store Visualization',\n", " scene=dict(xaxis_title='x',yaxis_title='y'),\n", " width=800,\n", " height=600,\n", " margin=dict(r=20, b=10, l=10, t=40)\n", " )\n", "\n", " fig.show()" ] }, { "cell_type": "code", "execution_count": null, "id": "2c250166-cb5b-4a75-8981-fae2d6dfe509", "metadata": {}, "outputs": [], "source": [ "show_embeddings_2d(result)" ] }, { "cell_type": "code", "execution_count": null, "id": "3b290e38-0800-4453-b664-7a7622ff5ed2", "metadata": {}, "outputs": [], "source": [ "def show_embeddings_3d(result):\n", " vectors = np.array(result['embeddings']) \n", " documents = result['documents']\n", " metadatas = result['metadatas']\n", " filenames = [metadata['filename'] for metadata in metadatas]\n", " filenames_unique = sorted(set(filenames))\n", "\n", " # color assignment\n", " color_map = {name: random_color() for name in filenames_unique}\n", " colors = [color_map[name] for name in filenames]\n", "\n", " tsne = TSNE(n_components=3, random_state=42)\n", " reduced_vectors = tsne.fit_transform(vectors)\n", "\n", " fig = go.Figure(data=[go.Scatter3d(\n", " x=reduced_vectors[:, 0],\n", " y=reduced_vectors[:, 1],\n", " z=reduced_vectors[:, 2],\n", " mode='markers',\n", " marker=dict(size=5, color=colors, opacity=0.8),\n", " text=[f\"Type: {t}
Text: {d[:100]}...\" for t, d in zip(filenames, documents)],\n", " hoverinfo='text'\n", " )])\n", "\n", " fig.update_layout(\n", " title='3D Chroma Vector Store Visualization',\n", " scene=dict(xaxis_title='x', yaxis_title='y', zaxis_title='z'),\n", " width=900,\n", " height=700,\n", " margin=dict(r=20, b=10, l=10, t=40)\n", " )\n", "\n", " fig.show()" ] }, { "cell_type": "code", "execution_count": null, "id": "45d1d034-2503-4176-b1e4-f248e31c4770", "metadata": {}, "outputs": [], "source": [ "show_embeddings_3d(result)" ] }, { "cell_type": "code", "execution_count": null, "id": "e79946a1-f93a-4b3a-8d19-deef40dec223", "metadata": {}, "outputs": [], "source": [ "# create a new Chat with OpenAI\n", "llm = ChatOpenAI(temperature=0.7, model_name=MODEL)\n", "\n", "# set up the conversation memory for the chat\n", "memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)\n", "\n", "# the retriever is an abstraction over the VectorStore that will be used during RAG\n", "retriever = vectorstore.as_retriever(search_kwargs={\"k\": 50})\n", "\n", "# putting it together: set up the conversation chain with the GPT 3.5 LLM, the vector store and memory\n", "conversation_chain = ConversationalRetrievalChain.from_llm(llm=llm, retriever=retriever, memory=memory)" ] }, { "cell_type": "code", "execution_count": null, "id": "59f90c85-c113-4482-8574-8a728ef25459", "metadata": {}, "outputs": [], "source": [ "def chat(question, history):\n", " result = conversation_chain.invoke({\"question\": question})\n", " return result[\"answer\"]" ] }, { "cell_type": "code", "execution_count": null, "id": "0520a8ff-01a4-4fa6-9dc8-57da87272edc", "metadata": {}, "outputs": [], "source": [ "view = gr.ChatInterface(chat, type=\"messages\").launch(inbrowser=True)" ] }, { "cell_type": "code", "execution_count": null, "id": "b4949b17-cd9c-4bff-bd5b-0f80df72e7dc", "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.11.11" } }, "nbformat": 4, "nbformat_minor": 5 }