{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "d3001150",
   "metadata": {
    "papermill": {
     "duration": 0.004023,
     "end_time": "2026-05-28T05:19:13.336520+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:19:13.332497+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# 30 · STORM — multi-perspective research → outline → article\n",
    "\n",
    "> **TL;DR.** 5-stage pipeline: brainstorm perspectives → questions per perspective → answer each → outline → write article section by section. Composes [Multi-Agent (nb 05)](./05_multi_agent.ipynb) + [Planning (nb 04)](./04_planning.ipynb).\n",
    "\n",
    "| Property | Value |\n",
    "|---|---|\n",
    "| Origin | Shao et al., *STORM* (Stanford 2024). [arXiv:2402.14207](https://arxiv.org/abs/2402.14207) |\n",
    "| Stages | perspectives → questions → answer → outline → write |\n",
    "| Cost | 1 + N + (N×K) + 1 + S where N=perspectives, K=questions/p, S=sections |\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "01b18795",
   "metadata": {
    "papermill": {
     "duration": 0.0,
     "end_time": "2026-05-28T05:19:13.340104+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:19:13.340104+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## 2 · Architecture\n",
    "\n",
    "```mermaid\n",
    "flowchart LR\n",
    "    A([topic]) --> P[PERSPECTIVES] --> Q[QUESTIONS] --> AN[ANSWER] --> O[OUTLINE] --> W[WRITE] --> Z([article])\n",
    "    style P fill:#e3f2fd,stroke:#1976d2\n",
    "    style W fill:#e8f5e9,stroke:#388e3c\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8d15e120",
   "metadata": {
    "papermill": {
     "duration": 0.008091,
     "end_time": "2026-05-28T05:19:13.348195+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:19:13.340104+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## 3 · Theory\n",
    "\n",
    "Plain article generation has one LLM \"write about X\" — produces shallow, single-perspective text. STORM forces breadth through *perspective diversity* and *atomic Q&A research* before writing.\n",
    "\n",
    "The deciding signals are all structured-output lists (perspectives, questions, sections). No numeric scoring → no flat-scoring risk."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e0243751",
   "metadata": {
    "papermill": {
     "duration": 0.0,
     "end_time": "2026-05-28T05:19:13.348195+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:19:13.348195+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## 4 · Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "d7f9c818",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-05-28T05:19:13.359025Z",
     "iopub.status.busy": "2026-05-28T05:19:13.359025Z",
     "iopub.status.idle": "2026-05-28T05:19:16.732192Z",
     "shell.execute_reply": "2026-05-28T05:19:16.731125Z"
    },
    "papermill": {
     "duration": 3.384979,
     "end_time": "2026-05-28T05:19:16.733174+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:19:13.348195+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">LLM: meta-llama/Llama-</span><span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">3.3</span><span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">-70B-Instruct  ·  Web research: Tavily (real)</span> <span style=\"color: #00ff00; text-decoration-color: #00ff00\">────────────────────────────────────────────</span>\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1;36mLLM: meta-llama/Llama-\u001b[0m\u001b[1;36m3.3\u001b[0m\u001b[1;36m-70B-Instruct  ·  Web research: Tavily \u001b[0m\u001b[1;36m(\u001b[0m\u001b[1;36mreal\u001b[0m\u001b[1;36m)\u001b[0m \u001b[92m────────────────────────────────────────────\u001b[0m\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from agentic_architectures import get_llm, enable_langsmith, settings\n",
    "from agentic_architectures.architectures import STORM\n",
    "from agentic_architectures.tools import web_search_tool\n",
    "from agentic_architectures.ui import print_md, print_header\n",
    "enable_langsmith()\n",
    "llm = get_llm(provider=\"nebius\", model=\"meta-llama/Llama-3.3-70B-Instruct\", temperature=0.4)\n",
    "\n",
    "# Real Tavily web search — every question in the ANSWER stage hits the live web.\n",
    "_tavily = web_search_tool(max_results=3)\n",
    "def web_search_fn(query: str) -> list[str]:\n",
    "    try:\n",
    "        result = _tavily.invoke(query)\n",
    "        if isinstance(result, list):\n",
    "            return [str(r.get('content', r))[:400] for r in result]\n",
    "        return [str(result)[:1000]]\n",
    "    except Exception as e:\n",
    "        return [f\"(web search unavailable: {e})\"]\n",
    "\n",
    "print_header(f\"LLM: {llm.model}  ·  Web research: Tavily (real)\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "019ade22",
   "metadata": {
    "papermill": {
     "duration": 0.004506,
     "end_time": "2026-05-28T05:19:16.743270+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:19:16.738764+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## 7 · Build the graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "9072992d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-05-28T05:19:16.753312Z",
     "iopub.status.busy": "2026-05-28T05:19:16.752276Z",
     "iopub.status.idle": "2026-05-28T05:19:17.428020Z",
     "shell.execute_reply": "2026-05-28T05:19:17.426091Z"
    },
    "papermill": {
     "duration": 0.682746,
     "end_time": "2026-05-28T05:19:17.430021+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:19:16.747275+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from IPython.display import Image, display\n",
    "arch = STORM(llm=llm, n_perspectives=3, questions_per_perspective=2, web_search_fn=web_search_fn)\n",
    "graph = arch.build()\n",
    "try: display(Image(graph.get_graph().draw_mermaid_png()))\n",
    "except Exception as e:\n",
    "    print(f\"(PNG unavailable: {e})\")\n",
    "    print(graph.get_graph().draw_mermaid())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aad6bab5",
   "metadata": {
    "papermill": {
     "duration": 0.013391,
     "end_time": "2026-05-28T05:19:17.447406+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:19:17.434015+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## 8 · Live run — article on agentic AI in 2024\n",
    "\n",
    "This will take 1-2 minutes (multiple LLM calls per stage)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "c480a9ae",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-05-28T05:19:17.467112Z",
     "iopub.status.busy": "2026-05-28T05:19:17.467112Z",
     "iopub.status.idle": "2026-05-28T05:20:21.884159Z",
     "shell.execute_reply": "2026-05-28T05:20:21.883145Z"
    },
    "papermill": {
     "duration": 64.428898,
     "end_time": "2026-05-28T05:20:21.886236+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:19:17.457338+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "N_PERSPECTIVES: 3\n",
      "N_QUESTIONS: 6\n",
      "N_SECTIONS: 4\n",
      "ARTICLE_CHARS: 4990\n",
      "\n",
      "=== PERSPECTIVES ===\n",
      "  [1] The rise of agentic AI architectures in 2024 marks a significant technical milestone, as it enables the development of more autonomous and adaptive systems.\n",
      "  [2] The increasing prevalence of agentic AI architectures in 2024 raises important social questions about the potential impact on employment and the distribution of wealth.\n",
      "  [3] The emergence of agentic AI architectures in 2024 can be seen as a natural progression of the historical trend towards increasingly sophisticated machine learning models, which has been driven by advances in computing power and data storage.\n",
      "\n",
      "=== QUESTIONS ===\n",
      "  [1] What are the key technical advancements that have enabled the development of agentic AI architectures in 2024?\n",
      "  [2] How can agentic AI architectures be designed to balance autonomy and adaptability with safety and control?\n",
      "  [3] How will the increasing use of agentic AI architectures in 2024 affect the job market and employment rates in various in\n",
      "  [4] What are the potential consequences of agentic AI architectures on the distribution of wealth, and how can their develop\n",
      "  [5] How do advances in computing power and data storage contribute to the development of agentic AI architectures?\n",
      "  [6] What role do historical trends in machine learning model sophistication play in shaping the emergence of agentic AI arch\n",
      "\n",
      "=== OUTLINE ===\n",
      "  ## Introduction to Agentic AI Architectures\n",
      "    - Definition of agentic AI architectures\n",
      "    - History and evolution of AI architectures\n",
      "    - Importance of agentic AI in 2024\n",
      "  ## Technical Advancements and Design Considerations\n",
      "    - Key technical advancements enabling agentic AI\n",
      "    - Mechanisms for balancing autonomy and safety\n",
      "    - Role of computing power and data storage in AI development\n",
      "  ## Societal Impacts and Regulatory Considerations\n",
      "    - Impact on job market and employment rates\n",
      "    - Consequences for wealth distribution and potential for income inequality\n",
      "    - Need for regulation and policies promoting transparency and fairness\n",
      "  ## Emergence and Future Directions\n",
      "    - Role of historical trends in machine learning model sophistication\n",
      "    - Potential applications and areas of research for agentic AI\n",
      "    - Future challenges and opportunities in agentic AI development\n",
      "\n",
      "=== ARTICLE (first 1500 chars) ===\n",
      "## Introduction to Agentic AI Architectures\n",
      "\n",
      "The year 2024 has witnessed a significant surge in the development and deployment of agentic AI architectures, marking a new era in artificial intelligence. At its core, agentic AI refers to autonomous systems that can perceive their environment, make decisions based on that perception, and take actions to achieve specific goals. This definition encapsulates the essence of agentic AI, highlighting its capacity for autonomy and adaptability. The history and evolution of AI architectures have paved the way for the emergence of agentic AI, with advancements in machine learning, computing power, and data storage playing pivotal roles. Understanding the importance of agentic AI in 2024 is crucial, as it promises to revolutionize various industries and aspects of life, from enhancing operational efficiency to potentially disrupting traditional employment structures. As we delve into the realm of agentic AI, it becomes imperative to consider how these architectures can be designed to balance autonomy with safety and control, ensuring that their development and integration into society are both beneficial and equitable.\n",
      "\n",
      "## Technical Advancements and Design Considerations\n",
      "\n",
      "The rise of agentic AI architectures in 2024 has been enabled by several key technical advancements. These include improvements in computing power and data storage, which have allowed for the development of more complex and sophisticated AI models. Additionally, advances\n"
     ]
    }
   ],
   "source": [
    "TOPIC = \"The rise of agentic AI architectures in 2024\"\n",
    "r = arch.run(TOPIC)\n",
    "print(f\"N_PERSPECTIVES: {r.metadata['n_perspectives']}\")\n",
    "print(f\"N_QUESTIONS: {r.metadata['n_questions']}\")\n",
    "print(f\"N_SECTIONS: {r.metadata['n_sections']}\")\n",
    "print(f\"ARTICLE_CHARS: {r.metadata['article_chars']}\")\n",
    "print()\n",
    "print(\"=== PERSPECTIVES ===\")\n",
    "for i, p in enumerate(r.metadata['perspectives'], 1):\n",
    "    print(f\"  [{i}] {p}\")\n",
    "print()\n",
    "print(\"=== QUESTIONS ===\")\n",
    "for i, q in enumerate(r.metadata['questions'], 1):\n",
    "    print(f\"  [{i}] {q['question'][:120]}\")\n",
    "print()\n",
    "print(\"=== OUTLINE ===\")\n",
    "for s in r.metadata['outline']:\n",
    "    print(f\"  ## {s['title']}\")\n",
    "    for kp in s['key_points'][:3]:\n",
    "        print(f\"    - {kp[:100]}\")\n",
    "print()\n",
    "print(\"=== ARTICLE (first 1500 chars) ===\")\n",
    "print(r.output[:1500])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5c737c4d",
   "metadata": {
    "papermill": {
     "duration": 0.004,
     "end_time": "2026-05-28T05:20:21.898202+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:20:21.894202+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## 9 · What we just observed\n",
    "\n",
    "The cells above ran STORM's 5-stage pipeline on the agentic-AI 2024 topic.\n",
    "\n",
    "### 9.1 · Pipeline statistics\n",
    "\n",
    "- **Perspectives generated**: 3\n",
    "- **Questions generated**: 6\n",
    "- **Outline sections**: 4\n",
    "- **Final article**: 4990 chars\n",
    "\n",
    "### 9.2 · The perspectives the agent chose\n",
    "\n",
    "| # | Perspective |\n",
    "|---|---|\n",
    "| 1 | The rise of agentic AI architectures in 2024 marks a significant technical milestone, as it enables the development of more autonomous and a… |\n",
    "| 2 | The increasing prevalence of agentic AI architectures in 2024 raises important social questions about the potential impact on employment and… |\n",
    "| 3 | The emergence of agentic AI architectures in 2024 can be seen as a natural progression of the historical trend towards increasingly sophisti… |\n",
    "\n",
    "### 9.3 · The outline that emerged\n",
    "\n",
    "| Title | # key points | First few points |\n",
    "|---|---|---|\n",
    "| `Introduction to Agentic AI Architectures` | 3 | Definition of agentic AI architectures, History and evolution of AI architectures, Importance of agentic AI in 2024 |\n",
    "| `Technical Advancements and Design Considerations` | 3 | Key technical advancements enabling agentic AI, Mechanisms for balancing autonomy and safety, Role of computing power and data storage in AI developme |\n",
    "| `Societal Impacts and Regulatory Considerations` | 3 | Impact on job market and employment rates, Consequences for wealth distribution and potential for income inequality, Need for regulation and policies  |\n",
    "| `Emergence and Future Directions` | 3 | Role of historical trends in machine learning model sophistication, Potential applications and areas of research for agentic AI, Future challenges and |\n",
    "\n",
    "### 9.4 · Patterns surfaced in this run\n",
    "\n",
    "- **✅ 3 distinct perspectives** generated.\n",
    "\n",
    "- **✅ 6 questions** ≈ 3 × questions_per_perspective.\n",
    "\n",
    "- **✅ Article assembled (4990 chars)** across 4 sections.\n",
    "\n",
    "### 9.5 · The takeaway\n",
    "\n",
    "STORM's value is in **diversity-by-construction**: a single \"write about X\" prompt collapses to one voice; STORM forces N perspectives → N×K questions → answers from many angles → an outline that reflects that breadth. Read § 9.2 to gauge whether the perspectives really diverge (paraphrases = pipeline broken) and § 9.3 to confirm the outline draws from multiple perspective threads."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "19550801",
   "metadata": {
    "papermill": {
     "duration": 0.002979,
     "end_time": "2026-05-28T05:20:21.905184+00:00",
     "exception": false,
     "start_time": "2026-05-28T05:20:21.902205+00:00",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## 11 · Failure modes & extensions\n",
    "\n",
    "| Failure | Mitigation |\n",
    "|---|---|\n",
    "| **Perspective overlap** | All N perspectives are paraphrases | Stricter schema; force \"must be substantively different from {prior}\" |\n",
    "| **Hallucinated answers** | LLM answering from no source | Add web search (pass `web_search_fn=`) |\n",
    "| **Section bloat** | Each section repeats prior section's content | Show prior sections to writer for de-dup |\n",
    "| **Cost** | 1+N+N*K+1+S calls — easily 12+ for default config | Cache; batch via asyncio |\n",
    "\n",
    "Extensions: (1) web-search backed answers, (2) iterative outline refinement, (3) per-perspective dedicated LLM (Llama for breadth, Qwen-Thinking for depth-sections).\n",
    "\n",
    "Reference: Shao et al., *STORM*. 2024. [arXiv:2402.14207](https://arxiv.org/abs/2402.14207)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.10.0"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 71.401164,
   "end_time": "2026-05-28T05:20:23.018671+00:00",
   "environment_variables": {},
   "exception": null,
   "input_path": "all-agentic-architectures/notebooks/30_storm.ipynb",
   "output_path": "all-agentic-architectures/notebooks/30_storm.ipynb",
   "parameters": {},
   "start_time": "2026-05-28T05:19:11.617507+00:00",
   "version": "2.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}