203_Rag_eval.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from langsmith.client import Client\n",
    "\n",
    "client = Client()\n",
    "dataset = client.clone_public_dataset(\n",
    "    \"https://smith.langchain.com/public/730d833b-74da-43e2-a614-4e2ca2502606/d\"\n",
    ")\n",
    "\n",
    "dataset_name = \"LCEL-QA\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "### INDEX\n",
    "\n",
    "from bs4 import BeautifulSoup as Soup\n",
    "from langchain_community.vectorstores import Chroma\n",
    "from langchain_openai import OpenAIEmbeddings\n",
    "from langchain_community.document_loaders.recursive_url_loader import RecursiveUrlLoader\n",
    "from langchain_text_splitters import RecursiveCharacterTextSplitter\n",
    "\n",
    "# Load docs\n",
    "url = \"https://python.langchain.com/v0.1/docs/expression_language/\"\n",
    "loader = RecursiveUrlLoader(\n",
    "    url=url, max_depth=20, extractor=lambda x: Soup(x, \"html.parser\").text\n",
    ")\n",
    "docs = loader.load()\n",
    "\n",
    "# Split into chunks\n",
    "text_splitter = RecursiveCharacterTextSplitter(chunk_size=4500, chunk_overlap=200)\n",
    "splits = text_splitter.split_documents(docs)\n",
    "\n",
    "# Embed and store in Chroma\n",
    "vectorstore = Chroma.from_documents(documents=splits, embedding=OpenAIEmbeddings())\n",
    "\n",
    "# Index\n",
    "retriever = vectorstore.as_retriever()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "### RAG bot\n",
    "\n",
    "import openai\n",
    "from langsmith import traceable\n",
    "from langsmith.wrappers import wrap_openai\n",
    "\n",
    "class RagBot:\n",
    "\n",
    "    def __init__(self, retriever, model: str = \"gpt-4o-mini\"):\n",
    "        self._retriever = retriever\n",
    "        # Wrapping the client instruments the LLM\n",
    "        self._client = wrap_openai(openai.Client())\n",
    "        self._model = model\n",
    "\n",
    "    @traceable()\n",
    "    def retrieve_docs(self, question):\n",
    "        return self._retriever.invoke(question)\n",
    "\n",
    "    @traceable()\n",
    "    def invoke_llm(self, question, docs):\n",
    "        response = self._client.chat.completions.create(\n",
    "            model=self._model,\n",
    "            messages=[\n",
    "                {\n",
    "                    \"role\": \"system\",\n",
    "                    \"content\": \"You are a helpful AI code assistant with expertise in LCEL.\"\n",
    "                    \" Use the following docs to produce a concise code solution to the user question.\\n\\n\"\n",
    "                    f\"## Docs\\n\\n{docs}\",\n",
    "                },\n",
    "                {\"role\": \"user\", \"content\": question},\n",
    "            ],\n",
    "        )\n",
    "\n",
    "        # Evaluators will expect \"answer\" and \"contexts\"\n",
    "        return {\n",
    "            \"answer\": response.choices[0].message.content,\n",
    "            \"contexts\": [str(doc) for doc in docs],\n",
    "        }\n",
    "\n",
    "    @traceable()\n",
    "    def get_answer(self, question: str):\n",
    "        docs = self.retrieve_docs(question)\n",
    "        return self.invoke_llm(question, docs)\n",
    "\n",
    "rag_bot = RagBot(retriever)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\"To build a Retrieval-Augmented Generation (RAG) chain using LCEL, you'd typically combine a retriever to fetch relevant documents with a language mode\""
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "response = rag_bot.get_answer(\"How to build a RAG chain in LCEL?\")\n",
    "response[\"answer\"][:150]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict_rag_answer(example: dict):\n",
    "    \"\"\"Use this for answer evaluation\"\"\"\n",
    "    response = rag_bot.get_answer(example[\"input_question\"])\n",
    "    return {\"answer\": response[\"answer\"]}\n",
    "\n",
    "def predict_rag_answer_with_context(example: dict):\n",
    "    \"\"\"Use this for evaluation of retrieved documents and hallucinations\"\"\"\n",
    "    response = rag_bot.get_answer(example[\"input_question\"])\n",
    "    return {\"answer\": response[\"answer\"], \"contexts\": response[\"contexts\"]}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain import hub\n",
    "from langchain_openai import ChatOpenAI\n",
    "\n",
    "# Grade prompt\n",
    "grade_prompt_answer_accuracy = prompt = hub.pull(\"langchain-ai/rag-answer-vs-reference\")\n",
    "\n",
    "def answer_evaluator(run, example) -> dict:\n",
    "    \"\"\"\n",
    "    A simple evaluator for RAG answer accuracy\n",
    "    \"\"\"\n",
    "\n",
    "    # Get question, ground truth answer, RAG chain answer\n",
    "    input_question = example.inputs[\"input_question\"]\n",
    "    reference = example.outputs[\"output_answer\"]\n",
    "    prediction = run.outputs[\"answer\"]\n",
    "\n",
    "    # LLM grader\n",
    "    llm = ChatOpenAI(model=\"gpt-4o-mini\", temperature=0)\n",
    "\n",
    "    # Structured prompt\n",
    "    answer_grader = grade_prompt_answer_accuracy | llm\n",
    "\n",
    "    # Run evaluator\n",
    "    score = answer_grader.invoke({\"question\": input_question,\n",
    "                                  \"correct_answer\": reference,\n",
    "                                  \"student_answer\": prediction})\n",
    "    score = score[\"Score\"]\n",
    "\n",
    "    return {\"key\": \"answer_v_reference_score\", \"score\": score}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "View the evaluation results for experiment: 'rag-answer-v-reference-f1df6e98' at:\n",
      "https://smith.langchain.com/o/eb86f40f-e90c-4eb8-846e-94f09f6daf2e/datasets/07ed7c14-cb07-4ba1-a086-f1f034d1dd01/compare?selectedSessions=e5899f21-3b46-4a19-8089-e20e8eb6e03a\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "860d28a4d77248e491075f44ab1a5361",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from langsmith.evaluation import evaluate\n",
    "\n",
    "experiment_results = evaluate(\n",
    "    predict_rag_answer,\n",
    "    data=dataset_name,\n",
    "    evaluators=[answer_evaluator],\n",
    "    experiment_prefix=\"rag-answer-v-reference\",\n",
    "    metadata={\"version\": \"LCEL context, gpt-4-0125-preview\"},\n",
    ")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.11.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}