⚛️ ACN v7.1 — Academic Chemistry Network

ACN / Academic Chemistry Network is an open-source local Gradio workbench for academic document parsing, concept extraction, concept-network construction, chemistry-inspired analogy analysis, visualization, and AI-assisted explanation.

ACN / 学术化学网络 是一个开源的本地 Gradio 工作台，用于学术文档解析、概念抽取、概念网络构建、化学启发式类比分析、可视化展示和 AI 辅助解释。

ACN does not claim to perform real chemical simulation. It uses chemistry-inspired metaphors and network analysis to observe academic literature and research-field structure.

ACN 不是 真正的化学仿真器。它使用化学启发式隐喻和网络分析来观察学术文献与研究领域结构。

---

What is ACN? / ACN 是什么？

ACN turns uploaded academic materials into a structured concept network. It can parse PDF, DOCX, XLSX, PPTX, and text files, extract academic or chemical concepts, build a graph of relationships, and apply several metaphorical analysis layers inspired by chemistry and physics.

ACN 会把上传的学术材料转化为结构化概念网络。它可以解析 PDF、DOCX、XLSX、PPTX 和文本文件，抽取学术或化学概念，构建概念关系图，并使用多种受化学与物理启发的隐喻分析层进行观察。

A simple way to understand it:

一个简单理解方式是：

Academic documents / 学术文档
        ↓
Text extraction / 文本提取
        ↓
Concept extraction / 概念抽取
        ↓
Concept network / 概念网络
        ↓
Chemistry-inspired analysis / 化学启发式分析
        ↓
Visualization, AI explanation, export / 可视化、AI解释、导出

---

Core Features / 核心功能

• Local Gradio web interface
• PDF / DOCX / XLSX / PPTX / TXT parsing
• Academic and chemical concept extraction
• Concept-network construction with NetworkX
• 2D and 3D visualizations with PyVis, Plotly, and local JS libraries
• Reaction-style analogy analysis
• Quantum-like and thermodynamic metaphor modules
• Catalysis and toxicity-style concept annotation
• Decay and lifecycle analysis for research concepts
• Counterfactual analysis of concept networks
• ARE: Academic Reaction Engine structural experiments
• ACL: Advanced Chemistry Layer annotations
• APL: OpenAI-compatible API backend
• Ollama local model backend
• Bilingual Chinese / English UI
• Export support for graph and report outputs

---

Why this project matters / 这个项目为什么有意思？

Many academic fields are not just collections of papers. They behave like evolving systems: concepts appear, react, dominate, decay, cluster, and sometimes become over-concentrated. ACN uses chemical and physical metaphors to help users observe this structure.

很多学术领域并不是简单的论文集合。它们更像不断演化的系统：概念会出现、反应、主导、衰减、聚类，有时也会过度集中。ACN 用化学和物理隐喻帮助用户观察这种结构。

It is useful for:

它适用于：

• Literature exploration / 文献探索
• Academic concept mapping / 学术概念制图
• Research-field structure analysis / 研究领域结构分析
• Interdisciplinary thinking / 跨学科思维
• Teaching and demonstration / 教学和演示
• AI-assisted academic explanation / AI 辅助学术解释
• Prototype research tools / 原型研究工具

---

Installation / 安装

Recommended Python version:

推荐 Python 版本：

Python 3.10 - 3.12

Install dependencies:

安装依赖：

pip install -r requirements.txt

Run ACN:

运行 ACN：

python run.py

Or directly:

或者直接运行：

python app.py

Windows:

start_acn.bat

Linux / macOS:

bash start_acn.sh

---

Optional AI Backends / 可选 AI 后端

ACN can run without an AI backend for basic parsing, extraction, network building, and visualization.

ACN 在没有 AI 后端时仍然可以进行基础解析、概念抽取、网络构建和可视化。

For AI-assisted explanations, ACN supports:

若要使用 AI 辅助解释，ACN 支持：

1. Ollama local backend / Ollama 本地后端
2. OpenAI-compatible API / OpenAI 兼容 API，也称 APL

APL can be configured through environment variables:

APL 可以通过环境变量配置：

APL_BASE_URL=https://api.openai.com/v1
APL_API_KEY=your_api_key
APL_MODEL=gpt-4o-mini

Or through acn_config.json based on acn_config.example.json.

也可以参考 acn_config.example.json 创建 acn_config.json。

---

Safety and Interpretation Boundary / 安全与解释边界

ACN is a research and visualization tool. It should not be treated as a chemical truth machine, a laboratory simulation engine, or a replacement for domain experts.

ACN 是研究和可视化工具，不应被当作化学真理机器、实验室仿真引擎，也不能替代领域专家判断。

Important boundaries:

重要边界：

• Reaction, quantum, toxicity, catalysis, and decay modules are metaphorical or heuristic unless explicitly connected to validated scientific data.
• Outputs should be interpreted as observation hints, not final scientific conclusions.
• AI explanations may hallucinate and must be checked by humans.
• Do not upload confidential, unpublished, copyrighted, or sensitive documents unless you understand the risks.

---

Repository Structure / 仓库结构

app.py                     Main Gradio UI / 主界面
run.py                     Safe launcher / 安全启动器
app_core.py                Application orchestration / 应用核心编排
concept_extractor.py        Concept extraction / 概念抽取
network_builder.py          Graph construction / 网络构建
visualizer.py               Visualization / 可视化
reaction_kinetics.py        Reaction-style models / 反应动力学隐喻
quantum_models.py           Quantum-like analysis / 类量子分析
catalysis_toxicity.py       Catalysis and toxicity annotation / 催化与毒性注释
counterfactual.py           Counterfactual experiments / 反事实实验
are_engine.py               Academic Reaction Engine / 学术反应引擎
acl_engine.py               Advanced Chemistry Layer / 高级化学层
ai_runtime.py               OpenAI-compatible runtime / APL 后端
ollama_client.py            Ollama backend / Ollama 后端
docs/                       Documentation / 文档
examples/                   Example inputs / 示例

---

Minimum Runtime Goal / 最低运行目标

ACN is considered minimally functional when it can:

当 ACN 能够做到以下几点时，可以认为达到最低可用状态：

• Launch the local Gradio interface
• Parse a small sample document
• Extract readable concepts
• Build a concept network
• Produce a 2D or 3D visualization
• Run at least one analysis module
• Export or display structured results
• Avoid garbled text, silent crashes, and unreadable errors

---

License / 许可证

This project is released under the MIT License. See LICENSE for details.

本项目使用 MIT License 开源。详情见 LICENSE。