An AI-powered research companion for machine learning practitioners — featuring research paper discovery, dataset auditing, and production-ready ML code generation.
NeuroGuide is an intelligent assistant designed to help ML practitioners navigate the complex landscape of machine learning research and implementation. Whether you're a researcher exploring new techniques, a startup building ML products, or a learner diving into the field, NeuroGuide provides contextual guidance throughout your ML journey.
┌─────────────────────────────────────────────────────────────────┐
│ Frontend (React + Vite) │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │
│ │ Dashboard │ │ Projects │ │ Chat │ │
│ └─────────────┘ └─────────────┘ └─────────────┘ │
└─────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ Backend │
│ ┌─────────────────────────────────────────────────────────┐ │
│ │ Multi-Agent Orchestration Layer │ │
│ │ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ │ │
│ │ │ Router │ │ Data │ │ Scholar │ │ Code │ │ │
│ │ │ Agent │ │ Engineer │ │ Agent │ │Generator │ │ │
│ │ └──────────┘ └──────────┘ └──────────┘ └──────────┘ │ │
│ │ ┌──────────┐ │ │
│ │ │ Critic │ │ │
│ │ │ Agent │ │ │
│ │ └──────────┘ │ │
│ └─────────────────────────────────────────────────────────┘ │
│ │ │
│ ┌───────────────────────────┴───────────────────────────┐ │
│ │ Edge Functions │ │
│ │ • Chat API (streaming responses) │ │
│ │ • Agent orchestration │ │
│ │ • External API integrations │ │
│ └───────────────────────────────────────────────────────┘ │
│ │ │
│ ┌───────────────────────────┴───────────────────────────┐ │
│ │ PostgreSQL Database │ │
│ │ • User profiles & preferences │ │
│ │ • Projects & conversations │ │
│ │ • Message history with metadata │ │
│ │ • Row Level Security (RLS) policies │ │
│ └───────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────┘
The core intelligence of NeuroGuide is powered by a collaborative multi-agent architecture:
| Agent | Role | Description |
|---|---|---|
| Router Agent | Intent Classification | Analyzes user queries and routes them to the appropriate specialist agent |
| Data Engineer | Dataset Auditing | Detects issues like class imbalance, data leakage, missing values, and outliers |
| Scholar Agent | Research Discovery | Retrieves papers from Semantic Scholar, ArXiv, and Papers with Code |
| Code Generator | Implementation | Generates production-ready ML code with safety checks and best practices |
| Critic Agent | Quality Assurance | Reviews all agent outputs for accuracy, completeness, and quality |
- Claude Sonnet 4.5 — Complex reasoning and research tasks
- Claude Haiku 4 — Rapid intent routing and classification
The application uses PostgreSQL with the following schema:
Stores user information and preferences.
- id (uuid, PK)
- email (text, required)
- full_name (text)
- avatar_url (text)
- goal (text) -- 'research' | 'startup' | 'learning'
- onboarding_completed (boolean)
- created_at (timestamp)
- updated_at (timestamp)ML projects with lifecycle tracking.
- id (uuid, PK)
- user_id (uuid, FK → profiles)
- title (text, required)
- description (text)
- status (text) -- 'ideation' | 'data_audit' | 'modeling' | 'completed'
- health_score (integer, 0-100)
- metadata (jsonb)
- created_at (timestamp)
- updated_at (timestamp)Chat sessions linked to projects.
- id (uuid, PK)
- project_id (uuid, FK → projects)
- user_id (uuid, FK → profiles)
- title (text)
- created_at (timestamp)Individual messages with rich metadata.
- id (uuid, PK)
- conversation_id (uuid, FK → conversations)
- role (text) -- 'user' | 'assistant' | 'system'
- content (text)
- metadata (jsonb) -- citations, code blocks, data issues
- created_at (timestamp)All tables are protected with Row Level Security (RLS) policies ensuring users can only access their own data.
| Technology | Purpose |
|---|---|
| React 18 | UI framework |
| Vite | Build tool & dev server |
| TypeScript | Type safety |
| Tailwind CSS | Styling |
| shadcn/ui | Component library |
| React Router | Client-side routing |
| TanStack Query | Data fetching & caching |
| React Hook Form | Form management |
| Zod | Schema validation |
| Technology | Purpose |
|---|---|
| Supabase | Backend-as-a-Service |
| PostgreSQL | Primary database |
| Edge Functions (Deno) | Serverless API endpoints |
| Supabase Auth | Authentication |
| RLS Policies | Data security |
neuroguide/
├── public/ # Static assets
│ ├── og-image.png # Social sharing image
│ ├── robot.txt
│ ├── placeholder.svg
│ └── favicon.ico
├── src/
│ ├── components/
│ │ ├── chat/ # Chat interface components
│ │ ├── dashboard/ # Dashboard widgets
│ │ ├── layout/ # Layout components
│ │ ├── ui/ # shadcn/ui components
│ │ └──NavLink.tsx
│ ├── hooks/
│ │ ├── use-mobile.tsx # Responsive utilities
│ │ ├── use-toast.ts
│ │ ├── useAuth.tsx # Authentication hook
│ │ ├── useChat.ts # Chat functionality
│ │ └── useProjects.ts # Project CRUD operations
│ ├── integrations/
│ │ └── supabase/ # Supabase client & types
│ ├── pages/ # Route components
│ ├── types/ # TypeScript interfaces
│ ├── lib/ # Utility functions
│ ├── test/
│ ├── App.css
│ ├── App.tsx
│ ├── index.css
│ ├── main.tsx
│ └── vite-env.d.ts
├── supabase/
│ ├── functions/
│ │ └── chat/ # Chat Edge Function
│ ├──migrations/ # Database, sql files
│ └── config.toml # Supabase configuration
├── .env
├── .gitignore
├── README.md
├── bun.lockb
├── components.json
├── eslint.config.js
├── index.html
├── package-lock.json
├── package.json
├── postcss.config.js
├── tailwind.config.js
├── tsconfig.app.json
├── tsconfig.json
├── tsconfig.node.json
├── vercel.json
├── vite.config.ts
└── vitest.config.ts
- Node.js 18+
- npm or bun
# Clone the repository
git clone https://github.com/nabakrishna/neuroguide-ai.git
cd neuroguide-ai
# Install dependencies
npm install
# Start development server
npm run devCreate a .env file with the following variables:
VITE_SUPABASE_URL=""
VITE_SUPABASE_PUBLISHABLE_KEY=""
VITE_SUPABASE_PROJECT_ID=""POST /functions/v1/chat
Handles streaming chat responses with multi-agent orchestration.
Request:
{
"messages": [
{ "role": "user", "content": "How do I handle class imbalance?" }
]
}Response: Server-Sent Events (SSE) stream with delta content.
The app uses Supabase Auth with:
- Email/password authentication
- Session persistence
- Protected routes
- Automatic token refresh
- Search across Semantic Scholar, ArXiv, and Papers with Code
- Citation extraction and formatting
- Abstract summarization
Automatically detects:
⚠️ Class imbalance- 🔴 Data leakage
- ❓ Missing values
- 📊 Outliers
- 🔄 Duplicate rows
- Production-ready implementations
- Best practices and safety checks
- Multiple framework support (PyTorch, TensorFlow, scikit-learn)
Ideation → Data Audit → Modeling → Completed
With health score tracking (0-100) for each project.
# Run tests
npm run test
# Run tests with coverage
npm run test -- --coverageDemo live on: https://neuroguide-ai.vercel.app/