Codon

中文 | English

合成生物学 AI 辅助设计平台：用 6 步点选式流程，为火星表面、木卫二冰下海洋、切尔诺贝利反应堆等极端环境生成可解释的工程微生物概念方案。

Codon 将“为一个极端环境设计一株能生存并执行生态修复任务的工程微生物”拆成一条可操作流水线：环境定义、任务选择、底盘推荐、功能蛋白选择、基因编辑方案生成、ODE 动态仿真。平台结合本地知识库、真实生物数据库查询、LLM 解释与科学计算模拟，面向教学、科研原型和演示场景。

安全说明：本项目输出为概念设计与教学级模拟，不应直接作为湿实验、环境释放、临床、生物安全审批或生产部署依据。

目录

• 核心能力
• Designer 六步流程
• Analysis 蛋白结构工作台
• 项目结构
• 技术栈
• 快速开始
• 环境变量
• 验证命令
• 部署
• 数据与数据库
• 文档
• 许可
• English

核心能力

• 极端环境建模：内置 11 个极端环境预设，并支持 9 维环境向量自定义。
• 生态修复任务选择：内置 10 类任务，包括产氧固碳、固氮造土、重金属吸收、放射核素固定、酸中和、塑料降解、溢油修复、有机氯降解、磷酸盐释放和生物膜固土。
• 可解释底盘推荐：基于耐受度、可工程化等级和任务相关性进行 Pareto 多目标排序，输出 Top 候选底盘。
• 蛋白与数据库支撑：维护 46 种功能蛋白记录，并接入 UniProt、NCBI E-utils、KEGG REST 查询与本地 SQLite 缓存。
• LLM 编辑方案生成：生成含载体、启动子、密码子优化、代谢负担、参考文献和 Kill Switch 的工程化编辑方案。
• ODE 动态仿真：使用 scipy.solve_ivp 求解种群、环境目标和限制性养分的耦合动态，并通过 SSE 推送 200 步轨迹。
• 项目化 Designer：裸 /designer 可直接开始设计；首次有效输入后自动保存为草稿 Project / Design / DesignerSession。
• 持续方案报告：每一步确认后持续更新 DesignReport，支持 Markdown 快照导出。
• 蛋白结构分析页：/analysis 内置 9EYS PSI 膜蛋白超复合物演示模型，支持 Mol* mmCIF 结构视图、ChimeraX GLB 预览、表示法切换和科研 Dossier 信息面板。
• 沉浸式前端体验：React + Three.js DNA 粒子背景、动效路由、项目绑定 Designer 会话和中英文界面。

Designer 六步流程

步骤	前端组件	后端路由	核心逻辑
1. 环境定义	EnvironmentStep.tsx	POST /designer/sessions/{sid}/environment	9 维环境向量；从后端加载 11 个环境预设，失败时使用本地 fallback
2. 任务选择	MissionStep.tsx	POST /designer/sessions/{sid}/mission	从后端加载 10 类任务，提交后返回 Top-6 底盘候选
3. 底盘筛选	ChassisStep.tsx	POST /designer/sessions/{sid}/chassis	biotype_service.recommend_chassis() Pareto 排序
4. 蛋白选择	ProteinStep.tsx	POST /designer/sessions/{sid}/protein	protein_service.recommend_proteins() + UniProt/BRENDA 链接 + LLM 解释
5. 编辑方案	GeneEditStep.tsx	POST /designer/sessions/{sid}/edit-plan	edit_plan_service.generate_edit_plans() 生成 3-5 个候选方案
6. 动态仿真	SimulationStep.tsx	POST /designer/sessions/{sid}/simulate	simulation_service.simulate() 生成 ODE 轨迹并通过 SSE 推送
方案报告	DesignReportPreview.tsx	GET /designer/sessions/{sid}/report / POST /designer/sessions/{sid}/report/exports/markdown	持续生成报告章节并导出 Markdown 快照
Copilot 对话	CopilotPanel.tsx	POST /designer/sessions/{sid}/copilot	OpenAI-compatible LLM 生成结构化 action；未配置 key 时使用本地规则兜底

每一步支持回滚；回滚会同步清理下游旧选择、候选快照和报告章节。

项目恢复链路是 Project -> Design -> DesignerSession：

• 裸 /designer 是主入口，不要求用户先创建项目。
• 项目卡片打开 /projects/:projectId/designer，后端恢复该项目默认或最新 Design 的 DesignerSession。
• DesignerSessionState 会返回 project_id、design_id、项目名、方案名、候选快照、仿真结果和报告状态所需字段。
• DesignReport 从设计过程开始持续记录，不等六步全部完成后才生成。

Analysis 蛋白结构工作台

/analysis 是答辩演示用的分子结构分析页，当前默认载入 PDB 9EYS：

• 结构视图：通过 Mol* 加载 frontend/public/models/psi_9eys.mmcif，支持 Mixed、Cartoon、Surface、Ball-stick 表示法切换、拖拽旋转、缩放、自动旋转和视角重置。
• ChimeraX 预览：通过 Three.js 加载优化后的 frontend/public/models/psi_9eys_chimerax.glb，用于展示 ChimeraX 导出的渲染外观。
• Molecular Dossier：右侧信息栏展示 9EYS / pdb_00009eys、EMDB EMD-50063、2.01 Å 分辨率、cryo-EM 方法、assembly 指标、论文 DOI / PubMed / funding 等科研上下文。
• 前端契约：frontend/src/lib/protein-models.ts 目前写死一个 demo record；后续可替换成后端返回的 { id, proteinName, structureUrl, previewUrl, format } 数据。

当前模型资产直接放在前端 public 目录中，便于离线演示。后续如果建设蛋白结构数据库，应将 mmCIF / GLB / OBZ 等资源改为后端或对象存储按 protein id 提供。

项目结构

Codon/
├── frontend/                         # React 19 + TypeScript + Vite
│   ├── src/
│   │   ├── App.tsx                   # 路由、鉴权、全局 DNA 背景
│   │   ├── api/                      # auth / projects / designer API 客户端
│   │   ├── auth/                     # 登录态与 token 处理
│   │   ├── components/               # 通用组件、Designer step 组件与 Analysis viewer
│   │   ├── i18n/                     # zh / en 本地化
│   │   ├── lib/                      # motion、DNA 场景、蛋白模型元数据和滚动辅助
│   │   └── views/                    # Home、Projects、Designer、Chat、Analysis、Login
│   ├── public/models/                # 9EYS mmCIF 与 ChimeraX GLB 演示资产
│   ├── vite.config.ts
│   └── package.json
│
├── backend/                          # FastAPI + SQLAlchemy
│   ├── app/
│   │   ├── api/v1/                  # auth、projects、designer、agents、chat
│   │   ├── services/                # 推荐、蛋白、编辑方案、仿真、方案报告、数据库查询、LLM
│   │   ├── models/                  # SQLAlchemy ORM
│   │   ├── schemas/                 # Pydantic schemas
│   │   ├── data/                    # 环境、任务、底盘、蛋白知识库 JSON
│   │   ├── main.py                  # FastAPI app 与路由注册
│   │   └── config.py                # Settings 与 CORS 配置
│   ├── alembic/                     # 数据库迁移
│   ├── Dockerfile
│   ├── render-start.sh
│   └── requirements.txt
│
├── render.yaml                       # Render Blueprint
├── docker-compose.yml                # 本地 PostgreSQL / Redis
├── CONTRIBUTING.md                   # 贡献指南
├── NOTICE                            # 第三方资源署名
├── LICENSE                           # MIT License
└── README.md

技术栈

层	技术
前端	React 19、TypeScript 5.8、Vite 6、React Router 7、Tailwind CSS 4、motion、Three.js、Mol*、Axios
后端	Python 3.9+、FastAPI 0.109、SQLAlchemy 2.0、Alembic、Pydantic 2、Uvicorn
认证	JWT、python-jose、bcrypt、HTTP-only cookie / Bearer token
数据与任务	SQLite / PostgreSQL、Redis、Celery
科学计算	scipy、Biopython、cobra
外部数据	UniProt REST、NCBI E-utils、KEGG REST
LLM	OpenAI-compatible client，可配置中转 API
部署	Vercel 前端、Render 后端、Supabase PostgreSQL

快速开始

1. 后端

推荐使用 Python 3.12 或项目目标范围内的 Python 3.9+。下面以 uv 为例：

cd backend
uv venv venv --python python3.12
uv pip install -r requirements.txt --python venv/bin/python
cp .env.example .env
venv/bin/python -m uvicorn app.main:app --reload

如果不使用 uv：

cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python -m uvicorn app.main:app --reload

后端默认地址：

• Health check: http://127.0.0.1:8000/health
• API docs: http://127.0.0.1:8000/api/docs

2. 前端

cd frontend
npm install
cp .env.example .env
npm run dev

前端默认地址：http://127.0.0.1:3000

演示分析页：http://127.0.0.1:3000/analysis

3. Docker 可选

docker-compose up --build

环境变量

前端

frontend/.env：

VITE_API_URL=http://127.0.0.1:8000/api/v1

生产环境必须包含 /api/v1 前缀，例如：

VITE_API_URL=https://<your-backend-domain>/api/v1

当前生产后端：

VITE_API_URL=https://codon-backend.onrender.com/api/v1

前端代码会拒绝使用旧的 mars-backend-2au4.onrender.com 地址，避免 Vercel 上残留的旧环境变量把生产站点连回已废弃后端。

后端

backend/.env 最少需要：

DATABASE_URL=sqlite:///./codon.db
SECRET_KEY=replace-with-a-long-random-secret
LLM_API_KEY=
LLM_BASE_URL=https://api.openai.com/v1
LLM_MODEL=gpt-4o
CORS_ORIGINS=["http://127.0.0.1:3000","http://localhost:3000"]

LLM_API_KEY 可留空；Copilot 会使用本地规则兜底。生产部署时建议使用 PostgreSQL，并将 SECRET_KEY、LLM_API_KEY、数据库连接串放在平台环境变量中。

验证命令

cd frontend && npm run lint
cd frontend && npm run build
cd backend && venv/bin/python -c "from app.main import app; print(app.title)"
curl -fsS http://127.0.0.1:8000/health

Analysis 页前端 smoke test 应确认：

1. /analysis 可打开。
2. 默认 Mol* mmCIF 结构 viewer 能加载 9EYS。
3. Mixed / Cartoon / Surface / Ball-stick 表示法按钮可切换。
4. ChimeraX 预览 能切换到 GLB 预览。
5. Overview / Assembly / Paper Dossier tabs 可切换，且无横向溢出。

一次最小 Designer API smoke test 应覆盖：

1. 注册 / 登录用户。
2. 获取环境预设：GET /api/v1/designer/presets/environments，应返回 11 条。
3. 获取任务预设：GET /api/v1/designer/presets/missions，应返回 10 条。
4. 创建 Designer session。
5. 提交环境向量。
6. 提交任务并获取底盘候选。
7. 读取持续方案报告，确认已生成对应章节。
8. 从项目路由恢复 Designer session，确认候选和已选状态不丢失。

部署

目标	推荐平台	说明
前端	Vercel	设置 VITE_API_URL=https://<backend>/api/v1 后重新部署
后端	Render	使用根目录 render.yaml Blueprint 或手动 Web Service
数据库	Supabase PostgreSQL	Render 生产环境使用 Supabase Session Pooler 连接串
CORS	后端环境变量	当前生产值：CORS_ORIGINS=["https://www.marssynbioagent.cn","https://marssynbioagent.cn"]，并保留 Vercel preview 域名正则

详细步骤见 backend/RENDER_DEPLOY.md。

旧 Railway / Sealos 部署文件已从开源树移除；当前推荐生产路径是 Render + Supabase。

数据与数据库

表 / 数据源	说明
users	用户、密码哈希、登录态
projects	用户项目
designs	项目下可命名、可恢复的设计方案
designer_sessions	归属到 Design 的 Designer 六步流程状态与候选快照
design_reports	随 Designer 步骤持续更新的方案文稿
report_exports	方案报告的导出快照记录
agent_runs	Legacy agent 编排记录
project_artifacts / project_datasets / project_jobs / project_view_snapshots	项目运行时资源
gene_modules / simulations / exports	设计与仿真产物
backend/app/data/*.json	环境、任务、底盘、蛋白知识库
backend/gene_query_cache.db	UniProt / NCBI / KEGG 查询缓存，TTL 7 天
frontend/public/models/psi_9eys.mmcif	Analysis 页默认 9EYS mmCIF 结构资产
frontend/public/models/psi_9eys_chimerax.glb	Analysis 页 ChimeraX GLB 预览资产
frontend/src/lib/protein-models.ts	当前前端蛋白模型 demo record 与科研展示元数据

文档

• CONTRIBUTING.md：贡献、开发与验证指南
• SECURITY.md：安全报告与密钥处理说明
• backend/RENDER_DEPLOY.md：Render + Supabase 部署指南
• NOTICE：第三方资源署名
• LICENSE：MIT License

许可

本项目代码以 MIT License 开源。第三方视觉资源署名见 NOTICE。

---

English

中文 | English

Codon is an AI-assisted synthetic biology design platform that turns extreme-environment microbe design into a guided six-step workflow. It helps users define an environment, choose an ecological mission, rank chassis organisms, select functional proteins, generate gene-editing plans, and run an educational ODE simulation.

The platform is intended for education, research prototyping, and product demonstrations. It combines curated local knowledge bases, real biological database lookups, LLM-generated explanations, and numerical simulation.

Safety note: Codon produces conceptual designs and simplified simulations. It must not be used directly as wet-lab instructions, environmental release guidance, clinical advice, biosafety approval material, or production deployment evidence.

Table of Contents

• Key Capabilities
• Designer Six-Step Workflow
• Analysis Protein Structure Workstation
• Project Structure
• Technology Stack
• Quick Start
• Environment Variables
• Verification
• Deployment
• Data and Database
• Documentation
• License

Key Capabilities

• Extreme-environment modeling: 11 built-in presets plus a custom 9-dimensional environment vector.
• Ecological mission selection: 10 mission classes, including oxygen and carbon fixation, nitrogen fixation, heavy-metal uptake, radionuclide immobilization, acid neutralization, plastic degradation, oil-spill remediation, organochlorine degradation, phosphate release, and biofilm-based soil stabilization.
• Explainable chassis recommendation: Pareto-style ranking based on environmental tolerance, genetic tractability, and mission relevance.
• Functional protein support: 46 curated protein records with UniProt, NCBI E-utils, KEGG REST integrations and local SQLite caching.
• LLM-generated edit plans: Candidate plans include vectors, promoters, codon optimization notes, metabolic burden, references, and Kill Switch status.
• ODE simulation: scipy.solve_ivp models population, environment target, and nutrient dynamics, streaming a 200-step trajectory over SSE.
• Project-backed Designer: Bare /designer can start immediately; the first valid input auto-saves a draft Project / Design / DesignerSession.
• Live design reports: Each confirmed step updates a DesignReport, with Markdown snapshot export support.
• Protein structure analysis page: /analysis includes the 9EYS PSI membrane-protein supercomplex demo model with Mol* mmCIF structure viewing, ChimeraX GLB preview, representation switching, and a research-grade Molecular Dossier panel.
• Interactive frontend: React, Three.js DNA particle background, animated routes, project-bound Designer sessions, and bilingual UI.

Designer Six-Step Workflow

Step	Frontend Component	Backend Route	Core Logic
1. Environment	EnvironmentStep.tsx	POST /designer/sessions/{sid}/environment	9-dimensional vector; loads 11 environment presets from the backend with local fallback
2. Mission	MissionStep.tsx	POST /designer/sessions/{sid}/mission	Loads 10 mission presets and returns Top-6 chassis candidates
3. Chassis	ChassisStep.tsx	POST /designer/sessions/{sid}/chassis	biotype_service.recommend_chassis() Pareto ranking
4. Protein	ProteinStep.tsx	POST /designer/sessions/{sid}/protein	protein_service.recommend_proteins() with UniProt/BRENDA links and LLM explanation
5. Edit Plan	GeneEditStep.tsx	POST /designer/sessions/{sid}/edit-plan	edit_plan_service.generate_edit_plans() generates 3-5 candidate plans
6. Simulation	SimulationStep.tsx	POST /designer/sessions/{sid}/simulate	simulation_service.simulate() streams an ODE trajectory over SSE
Design Report	DesignReportPreview.tsx	GET /designer/sessions/{sid}/report / POST /designer/sessions/{sid}/report/exports/markdown	Keeps report sections current and exports Markdown snapshots
Copilot Chat	CopilotPanel.tsx	POST /designer/sessions/{sid}/copilot	OpenAI-compatible LLM returns structured actions, with local-rule fallback when no key is configured

Each step supports rollback. Rollback clears downstream selections, candidate snapshots, and report sections.

The project restore chain is Project -> Design -> DesignerSession:

• Bare /designer is the primary entry and does not require users to create a project first.
• Project cards open /projects/:projectId/designer; the backend restores the default or latest Design and its DesignerSession.
• DesignerSessionState returns project_id, design_id, project name, design name, candidate snapshots, simulation result, and fields needed by report state.
• DesignReport starts during the design process instead of waiting for all six steps to finish.

Analysis Protein Structure Workstation

/analysis is the molecular-structure analysis page used for the current demonstration. It currently loads PDB 9EYS by default:

• Structure view: Mol* loads frontend/public/models/psi_9eys.mmcif and supports Mixed, Cartoon, Surface, and Ball-stick representations, drag rotation, zoom, auto-rotate, and reset view.
• ChimeraX preview: Three.js loads the optimized frontend/public/models/psi_9eys_chimerax.glb as a fallback look preview for the ChimeraX-exported rendering asset.
• Molecular Dossier: The right panel presents 9EYS / pdb_00009eys, EMDB EMD-50063, 2.01 Å resolution, cryo-EM method, assembly metrics, paper DOI / PubMed / funding, and other research context.
• Frontend contract: frontend/src/lib/protein-models.ts currently contains a single demo record. It can later be replaced by a backend response shaped around { id, proteinName, structureUrl, previewUrl, format }.

The current structure assets are stored directly under the frontend public directory for offline demonstrations. A later protein-structure database should serve mmCIF / GLB / OBZ assets by protein id from the backend or object storage.

Project Structure

Codon/
├── frontend/                         # React 19 + TypeScript + Vite
│   ├── src/
│   │   ├── App.tsx                   # Routing, auth guard, global DNA background
│   │   ├── api/                      # auth / projects / designer API clients
│   │   ├── auth/                     # Auth state and token handling
│   │   ├── components/               # Shared components, Designer steps, and Analysis viewers
│   │   ├── i18n/                     # zh / en localization
│   │   ├── lib/                      # motion, DNA scenes, protein model metadata, scroll helpers
│   │   └── views/                    # Home, Projects, Designer, Chat, Analysis, Login
│   ├── public/models/                # 9EYS mmCIF and ChimeraX GLB demo assets
│   ├── vite.config.ts
│   └── package.json
│
├── backend/                          # FastAPI + SQLAlchemy
│   ├── app/
│   │   ├── api/v1/                  # auth, projects, designer, agents, chat
│   │   ├── services/                # recommendation, proteins, edit plans, simulation, design reports, DB lookup, LLM
│   │   ├── models/                  # SQLAlchemy ORM
│   │   ├── schemas/                 # Pydantic schemas
│   │   ├── data/                    # environment, mission, chassis, protein JSON knowledge bases
│   │   ├── main.py                  # FastAPI app and route registration
│   │   └── config.py                # Settings and CORS
│   ├── alembic/                     # Database migrations
│   ├── Dockerfile
│   ├── render-start.sh
│   └── requirements.txt
│
├── render.yaml                       # Render Blueprint
├── docker-compose.yml                # Local PostgreSQL / Redis
├── CONTRIBUTING.md                   # Contribution guide
├── NOTICE                            # Third-party asset attributions
├── LICENSE                           # MIT License
└── README.md

Technology Stack

Layer	Technologies
Frontend	React 19, TypeScript 5.8, Vite 6, React Router 7, Tailwind CSS 4, motion, Three.js, Mol*, Axios
Backend	Python 3.9+, FastAPI 0.109, SQLAlchemy 2.0, Alembic, Pydantic 2, Uvicorn
Auth	JWT, python-jose, bcrypt, HTTP-only cookie / Bearer token
Data and jobs	SQLite / PostgreSQL, Redis, Celery
Scientific computing	scipy, Biopython, cobra
External data	UniProt REST, NCBI E-utils, KEGG REST
LLM	OpenAI-compatible client with configurable gateway
Deployment	Vercel frontend, Render backend, Supabase PostgreSQL

Quick Start

1. Backend

Python 3.9+ is supported. Python 3.12 is recommended for local development.

cd backend
uv venv venv --python python3.12
uv pip install -r requirements.txt --python venv/bin/python
cp .env.example .env
venv/bin/python -m uvicorn app.main:app --reload

Without uv:

cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python -m uvicorn app.main:app --reload

Default backend URLs:

• Health check: http://127.0.0.1:8000/health
• API docs: http://127.0.0.1:8000/api/docs

2. Frontend

cd frontend
npm install
cp .env.example .env
npm run dev

Default frontend URL: http://127.0.0.1:3000

Demo analysis page: http://127.0.0.1:3000/analysis

3. Optional Docker

docker-compose up --build

Environment Variables

Frontend

frontend/.env:

VITE_API_URL=http://127.0.0.1:8000/api/v1

Production values must include the /api/v1 prefix:

VITE_API_URL=https://<your-backend-domain>/api/v1

Current production backend:

VITE_API_URL=https://codon-backend.onrender.com/api/v1

The frontend refuses the legacy mars-backend-2au4.onrender.com value so a stale Vercel environment variable cannot connect production traffic to the retired backend.

Backend

Minimum backend/.env:

DATABASE_URL=sqlite:///./codon.db
SECRET_KEY=replace-with-a-long-random-secret
LLM_API_KEY=
LLM_BASE_URL=https://api.openai.com/v1
LLM_MODEL=gpt-4o
CORS_ORIGINS=["http://127.0.0.1:3000","http://localhost:3000"]

LLM_API_KEY can be left empty; the Copilot falls back to local rules. For production, use PostgreSQL and store SECRET_KEY, LLM_API_KEY, and the database URL in the deployment platform environment.

Verification

cd frontend && npm run lint
cd frontend && npm run build
cd backend && venv/bin/python -c "from app.main import app; print(app.title)"
curl -fsS http://127.0.0.1:8000/health

The Analysis page frontend smoke test should confirm:

1. /analysis opens.
2. The default Mol* mmCIF structure viewer loads 9EYS.
3. Mixed / Cartoon / Surface / Ball-stick representation buttons switch views.
4. ChimeraX 预览 switches to the GLB preview.
5. Overview / Assembly / Paper Dossier tabs switch without horizontal overflow.

A minimal Designer API smoke test should cover:

1. Register / log in a user.
2. Fetch environment presets: GET /api/v1/designer/presets/environments, expected count: 11.
3. Fetch mission presets: GET /api/v1/designer/presets/missions, expected count: 10.
4. Create a Designer session.
5. Submit an environment vector.
6. Submit a mission and receive chassis candidates.
7. Read the live design report and confirm matching sections exist.
8. Restore the Designer from a project route and confirm candidates and selections persist.

Deployment

Target	Recommended Platform	Notes
Frontend	Vercel	Set VITE_API_URL=https://<backend>/api/v1 and redeploy
Backend	Render	Use the root render.yaml Blueprint or create a Web Service manually
Database	Supabase PostgreSQL	Use the Supabase Session Pooler URL for Render
CORS	Backend environment variable	Current production value: CORS_ORIGINS=["https://www.marssynbioagent.cn","https://marssynbioagent.cn"], with the Vercel preview-domain regex kept enabled

See backend/RENDER_DEPLOY.md for detailed deployment steps.

Legacy Railway and Sealos deployment files have been removed from the open-source tree. The currently recommended production path is Render + Supabase.

Data and Database

Table / Source	Description
users	Users, password hashes, auth state
projects	User projects
designs	Named, restorable design variants under projects
designer_sessions	Design-owned six-step Designer state and candidate snapshots
design_reports	Live design documents updated as Designer steps complete
report_exports	Export snapshots for design reports
agent_runs	Legacy agent orchestration records
project_artifacts / project_datasets / project_jobs / project_view_snapshots	Project runtime resources
gene_modules / simulations / exports	Design and simulation artifacts
backend/app/data/*.json	Environment, mission, chassis, protein knowledge bases
backend/gene_query_cache.db	UniProt / NCBI / KEGG query cache, 7-day TTL
frontend/public/models/psi_9eys.mmcif	Default 9EYS mmCIF structure asset for the Analysis page
frontend/public/models/psi_9eys_chimerax.glb	ChimeraX GLB preview asset for the Analysis page
frontend/src/lib/protein-models.ts	Current frontend protein model demo record and research metadata

Documentation

• CONTRIBUTING.md: contribution, development, and verification guide
• SECURITY.md: security reporting and secret-handling notes
• backend/RENDER_DEPLOY.md: Render + Supabase deployment guide
• NOTICE: third-party asset attributions
• LICENSE: MIT License

License

Code is released under the MIT License. Third-party visual asset attributions are listed in NOTICE.