{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "d69eaf8c",
   "metadata": {},
   "source": [
    "# Modelo Híbrido de LLM em Um Notebook\n",
    "\n",
    "Autor: Daniel Fonseca (convertido para Colab)\n",
    "Data: 15/03/2026\n",
    "\n",
    "Este notebook executa o projeto completo em um único arquivo (.ipynb) no Colab."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "36144eed",
   "metadata": {},
   "source": [
    "## Descrição do modelo\n",
    "\n",
    "Combina L1 (conceitos), L2 (juízos kantianos), Módulo silogismo + Hempel + Popper, L3 (paraconsistente), L4 (sintese russelliana) e L5 (geração textual)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5131c88f",
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "!python -m pip install --upgrade pip\n",
    "!python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118 || true\n",
    "!python -m pip install transformers accelerate datasets sentencepiece python-docx protobuf<=3.20.3 scipy numpy psutil duckduckgo-search pyyaml fastapi uvicorn[standard]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cbdca50f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "import sys\n",
    "project_root = Path.cwd()\n",
    "if str(project_root) not in sys.path:\n",
    "    sys.path.insert(0, str(project_root))\n",
    "print('Project root:', project_root)\n",
    "print('Arquivos principais:')\n",
    "for f in ['pipeline.py','config_loader.py','l1_concept_table.py','l2_kantian_judgments.py','syllogism_module.py','l3_paraconsistent.py','l4_synthesis.py','l4_russell_equivalence.py','knowledge_base.py','l5_generation.py']:\n",
    "    print(f, (project_root/f).exists())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5e1e14bf",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pipeline import HybridLLMPipeline\n",
    "from config_loader import load_config\n",
    "config = load_config()\n",
    "pipeline = HybridLLMPipeline(config=config, verbose=True)\n",
    "print('Pipeline instanciado.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b7a7e630",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompts=['A água a 35 graus está quente ou fria?','O que é verdade?','A IA pode ter consciência?']\n",
    "for p in prompts:\n",
    "    print('---')\n",
    "    print('Prompt:',p)\n",
    "    r=pipeline.process(p)\n",
    "    print(r.response)\n",
    "    print('v=',r.truth_value,'estado=',r.state,'certeza=',r.certainty,'contradicao=',r.contradiction)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e55bffa1",
   "metadata": {},
   "source": [
    "## REPL interativo\n",
    "Execute esta célula e digite perguntas."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7d938a1c",
   "metadata": {},
   "outputs": [],
   "source": [
    "while True:\n",
    "    q = input('Prompt (sair para encerrar): ').strip()\n",
    "    if not q: continue\n",
    "    if q.lower() in {'sair','exit','quit'}:\n",
    "        print('Fim do REPL')\n",
    "        break\n",
    "    r=pipeline.process(q)\n",
    "    print('Resposta:',r.response)\n",
    "    print('v=',r.truth_value,'estado=',r.state,'certeza=',r.certainty,'contradicao=',r.contradiction)"
   ]
  }
 ],
 "metadata": {
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}