Skip to content
BranchesTags

Repository files navigation

ResearchMind · 智能科研助手

Your AI Partner in Materials Science

English | 中文版

📖 概述 (Overview)

ResearchMind 是一款基于多智能体架构的AI研究助手，专为材料科学研究领域深度定制。应用集成了四大核心功能模块，为研究人员提供从文献调研、数据查询、仿真计算到实验规划的一站式科研解决方案。

ResearchMind is an AI research assistant built on a multi-agent architecture, specifically tailored for the field of materials science. It integrates four core functional modules to provide researchers with a one-stop solution for everything from literature review and data queries to simulation and experimental planning.

✨ 核心优势 (Core Advantages)

  • 🧠 深度智能 (Deep Intelligence): 结合深度学习与大型语言模型，能够理解复杂的科研问题，并提供精准的分析与解答。
  • 🧩 模块化设计 (Modular Design): 灵活的多智能体架构，可根据研究需求调用最合适的智能体，精准满足不同研究场景。
  • ⚡ 高效执行 (Efficient Execution): 自动规划并并行执行复杂的科研任务，从文献检索到数据分析，大幅提升科研效率。
  • 🌐 多源整合 (Multi-Source Integration): 无缝对接多个权威学术数据库和计算平台，打破信息孤岛，提供全面的数据支持。
  • 🧠 Deep Intelligence: Combines deep learning and large language models to understand complex scientific problems, providing precise analysis and answers.
  • 🧩 Modular Design: A flexible multi-agent architecture that invokes the most suitable agent for your research needs, perfectly fitting various scenarios.
  • ⚡ Efficient Execution: Automatically plans and executes complex research tasks in parallel, from literature searches to data analysis, dramatically boosting your research efficiency.
  • 🌐 Multi-Source Integration: Seamlessly connects with multiple authoritative academic databases and computing platforms, breaking down information silos to provide comprehensive data support.

🚀 系统架构 (System Architecture)

核心功能模块 (Core Functional Modules)

智能体 (Agent) 功能特色 (Key Features) 数据源 / 能力 (Data Sources / Capabilities)
📚 文献研究 (Literature Research) 智能论文搜索、多源信息整合、深度内容分析、生成综合报告。 数据源: ArXiv, Tavily, Semantic Scholar, etc.
📊 数据库查询 (Database Query) 跨库智能检索、结构与性质查询、数据比对分析。 数据库: Materials Project, OQMD, COD, AFLOW
🔬 仿真计算 (Simulation & Calculation) 上传结构文件进行计算、支持多种物理性质分析、提供专业的可视化。 计算能力: 热导率 (Thermal Conductivity), 能量属性 (Energy Properties), 声子谱 (Phonon Spectrum)
🧪 实验规划 (Experiment Planning) 综合多源信息，制定科学严谨的实验方案与验证路径。 能力: 方案生成, 风险评估, 多智能体协同 (Plan Generation, Risk Assessment, Synergy)

Agent Key Features Data Sources / Capabilities
📚 Literature Research Smart paper search, multi-source information integration, deep content analysis, and comprehensive report generation. Data Sources: ArXiv, Tavily, Semantic Scholar, etc.
📊 Database Query Cross-database smart retrieval, structure and property queries, and comparative data analysis. Databases: Materials Project, OQMD, COD, AFLOW
🔬 Simulation & Calculation Upload structure files for calculation, supports various physical property analyses, and provides professional visualizations. Capabilities: Thermal Conductivity, Energy Properties, Phonon Spectrum
🧪 Experiment Planning Synthesizes multi-source info to create rigorous experimental plans and verification paths. Capabilities: Plan Generation, Risk Assessment, Multi-Agent Collaboration

📖 最佳实践指南 (Best Practices Guide)

🔬 示例1：材料热导率研究 (Example 1: Researching Thermal Conductivity)

  • 选择功能: 仿真计算助手 (Simulation & Calculation Agent)
  • 操作流程:
    1. 在左侧功能区选择“仿真计算助手”。
    2. 上传您的 CIF 晶体结构文件。
    3. 在输入框中输入指令:

    "计算这个结构的热导率和声子谱"

  • 预期输出:
    • ✅ 精确计算: 获得材料的热导率精确计算结果。
    • ✅ 专业图像: 生成声子谱图像，并附带专业分析。
    • ✅ 实时可视化: 在右侧展示区实时查看晶体结构与生成的声子谱图。
  • Select Agent: Simulation & Calculation Agent
  • Workflow:
    1. Choose "Simulation & Calculation Agent" from the left-hand function panel.
    2. Upload your CIF crystal structure file.
    3. Enter the command in the input box:

    "Calculate the thermal conductivity and phonon spectrum for this structure"

  • Expected Output:
    • ✅ Precise Calculation: Get accurate thermal conductivity results for your material.
    • ✅ Professional Graphics: Generate a phonon spectrum image with expert analysis.
    • ✅ Live Visualization: View the crystal structure and the generated phonon spectrum in real-time in the right-side display area.

⚗️ 示例2：晶体结构查询 (Example 2: Querying Crystal Structures)

  • 选择功能: 数据库查询助手 (Database Query Agent)
  • 操作流程:
    1. 选择“数据库查询助手”。
    2. 输入您想查询的材料，例如:

    "查询NaCl的晶体结构"

  • 预期输出:
    • ✅ 智能检索: 智能体将自动轮询所有可用数据库 (Materials Project, OQMD, COD, AFLOW)。
    • ✅ 快速响应: 返回首个成功命中的查询结果，并展示其详细信息和3D结构。
  • Select Agent: Database Query Agent
  • Workflow:
    1. Choose "Database Query Agent".
    2. Enter the material you want to query, for example:

    "Find the crystal structure of NaCl"

  • Expected Output:
    • ✅ Smart Retrieval: The agent will automatically query all available databases (Materials Project, OQMD, COD, AFLOW).
    • ✅ Fast Response: It will return the first successful match, displaying its detailed information and 3D structure.

📜 示例3：学术论文调研 (Example 3: Conducting a Literature Review)

  • 选择功能: 文献研究助手 (Literature Research Agent)
  • 操作流程:
    1. 选择“文献研究助手”。
    2. 输入您的研究主题，例如:

    "生成一份关于大语言模型在材料科学中应用的详细报告"

  • 预期输出:
    • ✅ 并行检索: 同时从 ArXiv, Tavily 等多个来源检索相关文献。
    • ✅ 深度分析: 对每篇关键论文进行深度剖析，提取核心观点和方法。
    • ✅ 综合报告: 生成一份结构清晰、内容详实的综合报告。
    • ✅ 数据文件: 提供一个包含所有引用文献、摘要和链接的 CSV 文件，方便您进一步分析。
  • Select Agent: Literature Research Agent
  • Workflow:
    1. Choose "Literature Research Agent".
    2. Enter your research topic, for example:

    "Generate a detailed report on the application of large language models in materials science"

  • Expected Output:
    • ✅ Parallel Search: Simultaneously retrieves literature from multiple sources like ArXiv and Tavily.
    • ✅ In-depth Analysis: Conducts a deep analysis of each key paper, extracting core ideas and methodologies.
    • ✅ Comprehensive Report: Generates a well-structured and detailed summary report.
    • ✅ Data File: Provides a CSV file containing all cited literature, abstracts, and links for your further analysis.

🧪 示例4：实验方案规划 (Example 4: Experimental Plan Planning)

  • 选择功能: 实验规划智能体 (Experiment Plan Agent)
  • 操作流程:
    1. 选择“实验规划智能体”。
    2. 输入您的研究目标，例如:

    "设计一个提高石墨烯/环氧树脂复合材料热导率的实验方案"

  • 预期输出:
    1. ✅ 综合调研: 自动调用文献和数据库智能体获取背景信息。
    2. ✅ 方案生成: 输出包含材料制备、表征测试、数据记录的完整实验流程。
    3. ✅ 风险评估: 提示潜在的实验风险并给出应对策略。
  • Select Agent: Experiment Plan Agent
  • Workflow:
    1. Choose "Experiment Plan Agent".
    2. Enter your research goal, for example:

    "Design an experimental plan to improve the thermal conductivity of graphene/epoxy composites"

  • Expected Output:
    1. ✅ Comprehensive Survey: Automatically calls literature and database agents for background info.
    2. ✅ Plan Generation: Outputs a complete experimental flow covering preparation, characterization, and data recording.
    3. ✅ Risk Assessment: Highlights potential risks and suggests mitigation strategies.

💰 收费标准 (Billing)

透明计费，无隐藏费用 (Transparent and Fair Billing)

  • 按工具调用计费: 只有当智能体在执行任务时调用外部API或进行复杂计算时才会产生费用。
  • 详细说明: 进入应用后，将有关于各项工具收费的详细说明。

Pay-per-use, no hidden fees.

  • Tool-Call Based: Charges only apply when an agent calls external APIs or performs complex calculations to complete a task.
  • Detailed Pricing: Detailed information about the cost of each tool is available within the application.

🎓 学术引用 (Academic Citation)

如果您在研究中使用了本应用的热导率计算功能，请引用以下论文。

If you use the thermal conductivity calculation feature of this application in your research, please cite the following paper.

BibTeX 格式:

@article{Liu2025PINK,
  author  = {Liu, Yujie and Wang, Xiaoying and Gao, Zhibin},
  title   = {{PINK: physical-informed machine learning for lattice thermal conductivity}},
  journal = {Journal of Materials Informatics},
  year    = {2025},
  volume  = {5},
  pages   = {12},
  doi     = {10.20517/jmi.2024.86}
}

BibTeX 格式:

@article{Liu2025PINK,
  author  = {Liu, Yujie and Wang, Xiaoying and Gao, Zhibin},
  title   = {{PINK: physical-informed machine learning for lattice thermal conductivity}},
  journal = {Journal of Materials Informatics},
  year    = {2025},
  volume  = {5},
  pages   = {12},
  doi     = {10.20517/jmi.2024.86}
}

👨‍💻 开发者指南 (Developer Guide)

🚀 快速启动 (Quick Start)

Linux / macOS / Git Bash

./start_linux.sh

Windows PowerShell

bash start_linux.sh

📋 系统要求 (System Requirements)

  • Python 3.10+
  • Node.js 18+
  • uv (Python package manager)

安装uv (Install uv)

# Linux/Mac
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows PowerShell
irm https://astral.sh/uv/install.ps1 | iex

🌐 访问地址 (Access URLs)

启动后可通过以下地址访问：

  • 前端界面 (Frontend UI): http://localhost:50001
  • 外部访问 (External Access): http://your-ip:50001
  • API文档 (API Docs): http://your-ip:50002/docs
  • 后端API (Backend API): http://localhost:50002 (local) or http://your-ip:50002 (external)

🔧 配置 (Configuration)

启动脚本会在项目根目录生成 .env 文件，下面列出需要关注的关键项：

API 密钥 (API Keys)

GOOGLE_API_KEY=your_google_api_key
DEEPSEEK_API_KEY=your_deepseek_api_key
TAVILY_API_KEY=your_tavily_api_key
MP_API_KEY=your_materials_project_api_key

网络相关变量 (Network Variables)

# UI 服务监听（start_complete.* 会把 50001 转发到该端口）
VITE_FRONTEND_HOST=127.0.0.1
VITE_FRONTEND_PORT=50010

# 前端调用 API/WS 时使用的相对路径 + 对外端口
VITE_API_URL=/api
VITE_API_PORT=50001
VITE_WS_URL=/ws
VITE_WS_PORT=50001

# 后端实际监听（默认仅限本机）
RESEARCHMIND_HTTP_HOST=127.0.0.1
RESEARCHMIND_HTTP_PORT=50002
RESEARCHMIND_WS_HOST=127.0.0.1
RESEARCHMIND_WS_PORT=50003

# MCP 服务
PAPER_SEARCH_MCP_HOST=127.0.0.1
PAPER_SEARCH_MCP_PORT=50004
SIMULATION_MCP_HOST=127.0.0.1
SIMULATION_MCP_PORT=50005
DATABASE_MCP_HOST=127.0.0.1
DATABASE_MCP_PORT=50006

提示 (Tip)

  • 如果不使用反向代理，可将 VITE_API_URL 改为 http://127.0.0.1:50002/api、VITE_WS_URL 改为 ws://127.0.0.1:50003/ws，并同步调整 VITE_API_PORT 与 VITE_WS_PORT。
  • 对外发布时，把 *_HOST 设置为 0.0.0.0，其它保持不变即可，由代理补全域名。

🛑 停止服务 (Stop Services)

按 Ctrl+C 停止所有服务，或使用：

./stop_linux.sh

✨ 启动脚本特性 (Start Script Features)

🔧 自动化功能 (Automation)

  • ✅ 环境检查 (Environment Check) - 自动检测uv、npm、Python等依赖
  • ✅ IP地址检测 (IP Detection) - 自动获取本机IP地址
  • ✅ 端口冲突处理 (Port Conflict Handling) - 检测并处理端口占用
  • ✅ 配置文件生成 (Config Generation) - 自动创建正确的.env配置
  • ✅ 服务健康检查 (Health Check) - 验证服务启动状态
  • ✅ 防火墙配置 (Firewall Config) - 自动添加防火墙规则（Linux）
  • ✅ 服务监控 (Service Monitoring) - 实时监控服务运行状态

🎯 智能特性 (Smart Features)

  • 🔄 强制重启 (Force Restart) - 自动停止现有服务并重启
  • 🌐 外部访问 (External Access) - 正确配置外部IP访问
  • 📊 状态显示 (Status Display) - 清晰显示所有服务状态和访问地址
  • 🚀 一键启动 (One-Click Start) - 无需手动配置，一键完成所有设置

📝 日志文件 (Log Files)

  • logs/backend.log – WebSocket/HTTP 主服务
  • logs/paper_search.log – 论文检索 MCP
  • logs/simulation.log – 仿真 MCP
  • logs/database.log – 数据库 MCP
  • logs/frontend.log – 前端构建/运行

🛠️ 故障排除 (Troubleshooting)

1. 前端依赖加载错误 (ERR_CONTENT_LENGTH_MISMATCH)

如果在访问前端时遇到 ERR_CONTENT_LENGTH_MISMATCH 错误，通常是由于 Nginx 缓冲设置导致的。系统已自动配置了解决方案：

  1. 确保使用了最新版本的 Nginx 配置文件
  2. 重启 Nginx 服务使配置生效： 
    # Windows
nginx -s reload

# Linux
sudo systemctl reload nginx

2. 端口冲突 (Port Conflict)

如果遇到端口冲突错误，可以修改 .env 文件中的端口配置。

About

An agent helps materials scientists quickly validate ideas with literature-assisted and experimental design

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages