DataBayt.AI Studio

Self-hosted · AI-assisted · Team-ready

A professional, fully self-hosted data annotation platform — built for teams who want full ownership of their labeling pipeline, AI-assisted workflows, and quality assurance tooling.

Why DataBayt.AI Studio?

Annotation tools shouldn't lock your data into a cloud you don't control.

DataBayt.AI Studio gives your team a complete annotation environment — with AI-assisted labeling, inter-annotator agreement tracking, role-based access, and bilingual (Arabic/English) support — running entirely on your own infrastructure.

Feature Highlights

Annotation Workspace

| Capability | Details | | --- | --- | | Upload formats | JSON, CSV, TXT, Hugging Face datasets | | Annotation modes | Record view · List view with filters | | AI assistance | Accept / Edit / Reject loop with confidence scores | | Custom forms | XML-based config or visual drag-and-drop builder | | Keyboard shortcuts | Full shortcut set + in-app reference panel | | Comments | Threaded comments per data point | | Undo / Redo | Full edit history within a session |

AI & Model Management

• Connect any provider: OpenAI, Anthropic, OpenRouter, SambaNova, Ollama (local)
• Build model profiles (prompt, temperature, max tokens, pricing)
• Set per-project allowed/default profiles via model policies
• Token estimation and cost-aware batch processing
• Hugging Face dataset import and private repo publishing

Team Collaboration

• Role-based access: admin · manager · annotator
• Invite-link onboarding with expiry, max-use limits, and role presets
• In-app notifications with deep-link navigation
• Audit log for uploads, AI runs, exports, and assignments
• Version history — snapshot and restore any project state
• Annotation guidelines — rich Markdown, always one click away

Quality Assurance

• Inter-Annotator Agreement (IAA) — configurable sampling rate and annotator count
• IAA Dashboard — per-item agreement scores with adjustable threshold slider
• Annotator Stats — speed, edit rate, rejection rate, agreement breakdown
• Quality charts — speed-by-annotator bar chart and per-annotator breakdown table

Bilingual Interface

• Full Arabic and English UI — every string, label, and message
• RTL layout switching with Cairo font for crisp Arabic rendering
• Language preference persisted in localStorage
• Relative dates localized (date-fns Arabic locale)

Security

• JWT authentication (8 h expiry, stored in sessionStorage)
• bcrypt password hashing (rounds = 12), transparent migration from legacy plaintext
• helmet security headers + login rate limiting (10 req / 15 min)
• API keys masked in all client responses
• CORS configurable via ALLOWED_ORIGINS

Tech Stack

| Layer | Technology | | --- | --- | | Frontend | React 18 · TypeScript · Vite · Tailwind CSS · shadcn/ui | | Backend | Express 5 (ESM) · Node.js ≥ 18 | | Database | SQLite via better-sqlite3 (WAL mode) | | Auth | jsonwebtoken · bcryptjs | | AI Providers | OpenAI · Anthropic · OpenRouter · SambaNova · Ollama | | i18n | i18next · react-i18next · i18next-browser-languagedetector | | Extras | driver.js · js-tiktoken · @huggingface/hub |

Quick Start

Prerequisites

• Node.js 18+

1 — Install dependencies

npm install

2 — Configure environment

cp .env.example .env

# Required
JWT_SECRET=your-random-secret-at-least-32-chars

# Optional
PORT=3000
DATA_DIR=server/data
ALLOWED_ORIGINS=http://localhost:8080

The server will warn on startup if JWT_SECRET is missing. Always set it in production.

3 — Run

npm run dev:all

Frontend: http://localhost:8080 · Backend: http://localhost:3000

Default credentials

| Username | Password | Note | | --- | --- | --- | | admin | admin | Password change required on first login |

Guided Onboarding

Every new user gets:

• An interactive tutorial (dashboard + workspace) powered by driver.js — deferred until after forced password change
• A demo practice project (sentiment analysis, 10 pre-labeled samples) to explore the workspace immediately

Project Structure

src/
├── components/
│   ├── DataLabelingWorkspace.tsx   # Main annotation workspace
│   ├── FormBuilder.tsx             # Visual annotation form builder
│   ├── DynamicAnnotationForm.tsx   # Runtime form renderer
│   ├── NotificationBell.tsx        # In-app notifications
│   ├── VersionHistory.tsx          # Snapshot & restore
│   ├── TemplatePickerModal.tsx     # Built-in + custom templates
│   ├── qa/
│   │   ├── AnnotationQualityDashboard.tsx
│   │   └── IAADashboard.tsx
│   └── Tutorial/
│       ├── tourSteps.ts            # Dashboard & workspace tour steps
│       └── useTutorial.ts
├── pages/
│   ├── Dashboard.tsx               # Projects, user management, login
│   ├── ModelManagement.tsx         # Providers, profiles, policies
│   └── ProjectSettings.tsx
├── contexts/
│   ├── AuthContext.tsx
│   └── LanguageContext.tsx         # RTL/LTR + language switching
├── i18n/
│   ├── index.ts                    # i18next configuration
│   └── locales/
│       ├── en.json                 # English strings
│       └── ar.json                 # Arabic strings
└── services/
    ├── apiClient.ts
    ├── xmlConfigService.ts
    └── exportService.ts

server/
├── index.js                        # Express app + middleware
├── middleware/auth.js               # JWT verify, requireAuth, requireRole
├── routes/
│   ├── projects.js                 # Projects, data points, snapshots, audit
│   ├── users.js                    # Auth, user CRUD, invite tokens
│   └── models.js                   # Connections, profiles, policies
└── services/
    ├── database.js                 # SQLite schema, migrations, seed
    └── notificationService.js

Access Control

| Role | Capabilities | | --- | --- | | admin | Full access — users, all projects, model management | | manager | Manage assigned projects, annotator assignments, model management | | annotator | Access only assigned projects; annotate data |

Data Formats

| Format | Behavior | | --- | --- | | CSV | All columns preserved as metadata; choose display columns in workspace | | JSON | Flexible payloads — text and image-style records supported | | TXT | Each line becomes a separate annotation item | | Hugging Face | Browse and import public datasets directly from the workspace |

Deployment

# Build frontend
npm run build

# Start production server
npm start

• Serve the dist/ folder as static files from the same Express process or a reverse proxy (nginx).
• Set JWT_SECRET to a strong random value: openssl rand -hex 32
• SQLite database is created at server/data/databayt.sqlite — override with DATA_DIR.
• All provider API keys are stored server-side and never exposed in full to clients.

Troubleshooting

Provider / model list not loading

• Confirm backend is running: npm run dev:all
• Verify API key in Model Management → Connections
• Check browser Network tab for proxy route errors (/api/openai/*, etc.)

Upload issues

• Confirm file is valid JSON / CSV / TXT
• For CSV, ensure column headers are present in the first row

AI processing errors

• Verify the model profile has an active provider connection
• Check API key credits / rate limits
• For Ollama, confirm the endpoint is reachable: http://localhost:11434

Access denied on project or model pages

• Confirm the user's role and project assignment in User Management

401 after server restart

• sessionStorage tokens are tab-scoped and do not survive a browser session restart — simply log in again

License

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0).

You are free to use, modify, and distribute this software under the terms of the AGPL-3.0. Any modified version deployed as a network service must also be made available under the same license.

See the LICENSE file for full details.

Built with care by the DataBayt team.