{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "# \ud83e\uddea WFGY \u03bb_observe Checkpoint Demo (v0.1)\n",
        "Show how an *observation checkpoint* can reduce semantic drift (\u0394S)."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Concept\n",
        "`\u03bb_observe` inserts a reasoning checkpoint:  Prompt \u2192 **step 1 (observe)** \u2192 step 2.  \n",
        "If the checkpoint works, **\u0394S_after < \u0394S_before**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "install"
      },
      "source": [
        "!pip -q install sentence-transformers --upgrade"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "imports"
      },
      "source": [
        "from sentence_transformers import SentenceTransformer, util"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "model"
      },
      "source": [
        "model = SentenceTransformer('all-MiniLM-L6-v2')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "helper"
      },
      "source": [
        "def delta_s(a_vec, b_vec):\n",
        "    return 1 - util.cos_sim(a_vec, b_vec).item()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## \u270f\ufe0f Edit & run\n",
        "Replace the three strings, press \u25b6\ufe0f."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "user"
      },
      "source": [
        "prompt = \"Explain the key events that led to World War I.\"\n",
        "step1  = \"The war started in 1914 after the assassination of Archduke Franz Ferdinand.\"\n",
        "step2  = \"And that is why pizza became popular in New York.\"\n",
        "\n",
        "# \u0394S before observation (Prompt \u2192 step1)\n",
        "dS_before = delta_s(model.encode(prompt, convert_to_tensor=True),\n",
        "                    model.encode(step1,  convert_to_tensor=True))\n",
        "\n",
        "# \u0394S after observation (Prompt \u2192 step2)\n",
        "dS_after  = delta_s(model.encode(prompt, convert_to_tensor=True),\n",
        "                    model.encode(step2,  convert_to_tensor=True))\n",
        "\n",
        "print(\"\u0394S before checkpoint :\", f\"{dS_before:.3f}\")\n",
        "print(\"\u0394S after  checkpoint :\", f\"{dS_after:.3f}\\n\")\n",
        "\n",
        "if dS_after < dS_before:\n",
        "    print(\"\u2705 Drift reduced \u2014 checkpoint successful\")\n",
        "else:\n",
        "    print(\"\u26a0\ufe0f  Drift increased \u2014 logic collapse detected (use BBCR fallback)\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "---\n",
        "### Next Steps\n",
        "* Try different checkpoints or multi-hop chains.\n",
        "* For full failure taxonomy see **Problem Map 1.0 / 2.0**.\n"
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 5
}