{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "ebeba428",
   "metadata": {},
   "source": [
    "# ✅ RAG JuJutsu PoC (Notebook with Joblib, FAISS, ChatGPT API)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8bdfd3c8",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "\n",
    "!pip install --quiet openai langchain faiss-cpu PyPDF2 sentence-transformers joblib\n",
    "!pip install ipywidgets==7.7.2\n",
    "!jupyter nbextension enable --py widgetsnbextension\n",
    "!jupyter notebook\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "49ee7721",
   "metadata": {},
   "outputs": [],
   "source": [
    "from PyPDF2 import PdfReader\n",
    "from langchain.text_splitter import RecursiveCharacterTextSplitter\n",
    "\n",
    "def load_pdf_chunks(pdf_path):\n",
    "    reader = PdfReader(pdf_path)\n",
    "    raw_text = \"\"\n",
    "    for page in reader.pages:\n",
    "        raw_text += page.extract_text() + \"\\n\"\n",
    "\n",
    "    splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)\n",
    "    return splitter.split_text(raw_text)\n",
    "\n",
    "chunks = load_pdf_chunks(\"JuJutsu-Contexto-Significado-Conexiones-Historia.pdf\")\n",
    "print(f\"Loaded {len(chunks)} chunks\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8109b626-0179-43e2-b924-65afe9af1e4e",
   "metadata": {},
   "outputs": [],
   "source": [
    "import openai"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "371c637e",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "import faiss\n",
    "import numpy as np\n",
    "import joblib\n",
    "\n",
    "def get_openai_embeddings(texts):\n",
    "    embeddings = []\n",
    "    for text in texts:\n",
    "        response = openai.Embedding.create(\n",
    "            model=\"text-embedding-3-small\",\n",
    "            input=text\n",
    "        )\n",
    "        vector = response['data'][0]['embedding']\n",
    "        embeddings.append(vector)\n",
    "    return np.array(embeddings)\n",
    "\n",
    "embeddings = get_openai_embeddings(chunks)\n",
    "index = faiss.IndexFlatL2(embeddings.shape[1])\n",
    "index.add(np.array(embeddings))\n",
    "\n",
    "joblib.dump((chunks, index), \"rag_model.joblib\")\n",
    "print(\"Chunks and index serialized to rag_model.joblib\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "28ce4963",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "import joblib\n",
    "chunks, index = joblib.load(\"rag_model.joblib\")\n",
    "print(\"Chunks and index loaded from rag_model.joblib\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "51a89e77",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "def search(query, k=3):\n",
    "    response = openai.Embedding.create(\n",
    "        model=\"text-embedding-3-small\",\n",
    "        input=query\n",
    "    )\n",
    "    query_vec = np.array([response['data'][0]['embedding']])\n",
    "    scores, indices = index.search(query_vec, k)\n",
    "    return [chunks[i] for i in indices[0]]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "34315775",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "import os\n",
    "import openai\n",
    "from openai import OpenAI\n",
    "\n",
    "os.environ[\"OPENAI_API_KEY\"] = os.getenv(\"OPENAI_API_KEY\")\n",
    "client = OpenAI()\n",
    "\n",
    "def chat_no_rag(question):\n",
    "    response = client.chat.completions.create(\n",
    "        model=\"gpt-3.5-turbo\",\n",
    "        messages=[\n",
    "            {\"role\": \"user\", \"content\": question}\n",
    "        ],\n",
    "        temperature=0.5,\n",
    "        max_tokens=200,\n",
    "    )\n",
    "    return response.choices[0].message.content\n",
    "\n",
    "def chat_with_rag(question, retrieved_chunks):\n",
    "    context = \"\\n\".join(retrieved_chunks)\n",
    "    prompt = f\"Usa el siguiente contexto para responder la pregunta:\\n\\n{context}\\n\\nPregunta: {question}\"\n",
    "\n",
    "    response = client.chat.completions.create(\n",
    "        model=\"gpt-3.5-turbo\",\n",
    "        messages=[\n",
    "            {\"role\": \"user\", \"content\": prompt}\n",
    "        ],\n",
    "        temperature=0.3,\n",
    "        max_tokens=200,\n",
    "    )\n",
    "    return response.choices[0].message.content\n",
    "\n",
    "def chat_with_rag_enhanced(question, retrieved_chunks):\n",
    "    context = \"\\n\".join(retrieved_chunks)\n",
    "    prompt = (\n",
    "        \"Eres un experto en historia marcial. \"\n",
    "        \"Usa el siguiente contexto histórico para responder con precisión y detalle.\\n\\n\"\n",
    "        f\"Contexto:\\n{context}\\n\\n\"\n",
    "        f\"Pregunta: {question}\\nRespuesta:\"\n",
    "    )\n",
    "\n",
    "    response = client.chat.completions.create(\n",
    "        model=\"gpt-3.5-turbo\",\n",
    "        messages=[\n",
    "            {\"role\": \"user\", \"content\": prompt}\n",
    "        ],\n",
    "        temperature=0.2,\n",
    "        max_tokens=200,\n",
    "    )\n",
    "    return response.choices[0].message.content\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "900dfdfa",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# Example query\n",
    "query = \"¿Cuál es el origen del JuJutsu en Japón?\"\n",
    "retrieved = search(query)\n",
    "\n",
    "print(\"🔹 Sin RAG:\")\n",
    "print(chat_no_rag(query))\n",
    "\n",
    "print(\"\\n🔹 Con RAG:\")\n",
    "print(chat_with_rag(query, retrieved))\n",
    "\n",
    "print(\"\\n🔹 Con RAG + Prompt mejorado:\")\n",
    "print(chat_with_rag_enhanced(query, retrieved))\n"
   ]
  }
 ],
 "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": "3.11.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}