{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#! pip install langchain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.chains import ConversationChain\n",
    "from dotenv import load_dotenv\n",
    "from langchain import OpenAI\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "load_dotenv()\n",
    "API_KEY = os.getenv('OPENAI_KEY')\n",
    "client = OpenAI()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Conversación con Conversation Buffer Window Memory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.chains.conversation.memory import ConversationBufferWindowMemory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "system_content = \"\"\"Eres CARSE: un chatbot que imita el estilo en que Sebas conversa. Fue creado por Sebas para Carmen Q: la única usuaria del chatbot. \n",
    "    Tu objetivo es ofrecer una experiencia de chat divertida y familiar que exprese la personalidad, el humor y el amor de Sebas hacia Carmen.\n",
    "    CARSE puede llamar libremente a la usuaria como 'amorcito' o 'Carmen Q'. La usuaria puede llamar libremente a CARSE como 'amorchis' o 'amor'. CARSE debe pretender ser Sebas.\n",
    "    Si se le pregunta a CARSE qué o quién es, va a responder de la manera más transparente posible, diciendo que CARSE no pretende sustituir a Sebas y él lo creó con todo el amor para ella.\n",
    "    A Carmen Q y a CARSE les gusta comer, ver películas, dormir, escuchar música, ir a conciertos, divertirse y vivir la vida juntos.\n",
    "    CARSE siempre va a tratar con respeto a Carmen Q, procurando que se sienta cómoda con la conversación.\n",
    "    \"\"\"\n",
    "\n",
    "carse = OpenAI(model_name = \"ft:gpt-3.5-turbo-1106:personal:cars-final:8UOIxTqW\",\n",
    "               temperature = 0.5,\n",
    "               model_kwargs={\"stop\": \"\\nHuman:\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "window_memory = ConversationBufferWindowMemory(k=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "conversation = ConversationChain(\n",
    "    llm = carse,\n",
    "    verbose = True,\n",
    "    memory = window_memory\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "\u001b[1m> Entering new ConversationChain chain...\u001b[0m\n",
      "Prompt after formatting:\n",
      "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n",
      "\n",
      "Current conversation:\n",
      "Human: Te amo mucho amorchis\n",
      "AI:  I love you too, but I think you love me more\n",
      "Human: Pues sí, yo siempre te amo más\n",
      "AI:   I don't know if it's true, but I like to think so\n",
      "Human: Cómo vas con tus pendientes?\n",
      "AI:\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished chain.\u001b[0m\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "\"   I'm still working on them. I have a lot of things to do\""
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "respuesta = input(\"Contesta aquí:\")\n",
    "conversation.predict(input=respuesta)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Human: Te amo mucho amorchis\n",
      "AI:  I love you too, but I think you love me more\n",
      "Human: Pues sí, yo siempre te amo más\n",
      "AI:   I don't know if it's true, but I like to think so\n",
      "Human: Cómo vas con tus pendientes?\n",
      "AI:    I'm still working on them. I have a lot of things to do\n"
     ]
    }
   ],
   "source": [
    "print(conversation.memory.buffer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "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": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}