DataGems

DataGems is a commercial-grade synthetic data generation project that leverages the Ratio1 SDK and on-edge-cloud sovereign infrastructure to help teams create high-quality structured datasets quickly, safely, and repeatedly.

It combines a Next.js full-stack app with Ratio1-backed persistence and strict authentication to support auditable AI-data workflows for product, research, and regulated environments.

Production deployment: https://datagems.app

Why DataGems

• Generate synthetic datasets with predictable quality using one model call per record.
• Keep full traceability of jobs, progress, outputs, and metrics in persistent storage.
• Enforce authentication and session control across all API and UI actions.
• Export usable results fast (JSON/CSV) for analytics, testing, and model development.
• Fit modern AI product teams that need delivery speed without compromising governance.

Ownership and Commercial Context

Owner:SmartClover SRL (Romania)

DataGems is part of SmartClover SRL's product strategy and aligns with the company's public objectives:

• Human-in-the-loop AI systems.
• Data sovereignty and controlled deployments ("your AI, your Data").
• Practical healthcare AI productization through SaaS/PaaS delivery models.

As published on SmartClover channels (accessed February 17, 2026), SmartClover operates a portfolio that includes:

• CerviGuard (MDR Class I cervical cancer screening companion app).
• Evidence-Linked Healthcare Research Platform.
• Digital Resilience Platform for Healthcare.
• Creative Education Experience Platform.

DataGems serves as an enabling layer for synthetic data operations across these product directions and adjacent enterprise use cases.

Quickstart

1. Install dependencies (Node 18+):

npm install

2. Set environment variables. For local mock mode (admin/admin), this is enough in .env.local:

DATAGEN_SESSION_SECRET=dev-session-secret-change-me
DATAGEN_MOCK_CSTORE=true
DATAGEN_MOCK_INFERENCE_API=true
R1EN_CHAINSTORE_PEERS=local
R1EN_HOST_ADDR=local

3. Run:

npm run dev

Environment variables

• EE_CHAINSTORE_API_HOST / EE_CHAINSTORE_API_PORT (required): Ratio1 CStore endpoint (fallback: EE_CHAINSTORE_API_URL).
• EE_R1FS_API_HOST / EE_R1FS_API_PORT (optional): Ratio1 file store (fallback: EE_R1FS_API_URL, not used yet).
• R1EN_CSTORE_AUTH_HKEY: Hash key for auth (fallback: EE_CSTORE_AUTH_HKEY).
• R1EN_CSTORE_AUTH_SECRET: Pepper for password hashing (fallback: EE_CSTORE_AUTH_SECRET).
• R1EN_CSTORE_AUTH_BOOTSTRAP_ADMIN_PWD: Bootstrap admin password (fallback: EE_CSTORE_AUTH_BOOTSTRAP_ADMIN_PWD, legacy: EE_CSTORE_AUTH_BOOTSTRAP_ADMIN_PW).
• DATAGEN_SESSION_SECRET: Secret used to sign session cookies/JWTs. In development/test, if omitted, DataGems uses an insecure built-in fallback secret for local runs only.
• DATAGEN_APP_HOST / DATAGEN_APP_PORT: Public app host/port (defaults: $R1EN_HOST_IP / 3000, fallback: DATAGEN_APP_URL).
• DATAGEN_INFERENCE_HOST / DATAGEN_INFERENCE_PORT: Inference gateway host/port (defaults: $R1EN_HOST_IP / $API_PORT, fallback: DATAGEN_INFERENCE_BASE_URL).
• DATAGEN_SMTP_HOST: SMTP host for signup email delivery (default smtp.resend.com).
• DATAGEN_SMTP_PORT: SMTP port (default 465).
• DATAGEN_SMTP_USER: SMTP username (default resend).
• DATAGEN_SMTP_PASS: SMTP password/API key (default empty; optional to define, required to actually send email).
• DATAGEN_SMTP_FROM: Sender email (default no-reply@datagems.app).
• R1EN_CHAINSTORE_PEERS: Peer list for multi-instance execution (comma-separated or JSON array string).
• R1EN_HOST_ADDR: Current instance peer id (must match one entry in R1EN_CHAINSTORE_PEERS).
• LOG_INFERENCE_REQUESTS: When true, logs outgoing inference requests (auth header redacted).
• DATAGEN_LOG_R1FS_CALLS: When true, logs R1FS upload/download start/success/error events.
• RETRY_INFERENCE_ON_FAILURE: When true, retries one extra inference call on failure/parse errors.
• NEXT_PUBLIC_SHOW_FAILURES: When true, shows failure counts in UI task cards.
• DATAGEN_MAX_RECORDS_PER_JOB: Max records per job (default 200).
• DATAGEN_MAX_EXTERNAL_API_CONFIGS: Max saved external API profiles per user (default 10).
• DATAGEN_MOCK_CSTORE: When true, uses in-memory mock CStore/auth (admin/admin, test_user/testtest). In development/test, DataGems auto-falls back to mock mode if auth/CStore env is missing.
• DATAGEN_MOCK_INFERENCE_API: When true, uses in-memory mock inference that returns random JSON records.
• DATAGEN_JOB_POLL_SECONDS: Worker poll interval for queued/running jobs (default 5).
• DATAGEN_UPDATE_EVERY_K_REQUESTS: Persist/update cadence during generation (default 5).
• DATAGEN_MAX_CONCURRENT_JOBS_PER_INSTANCE: Max jobs processed in parallel per instance (default 1).
• DATAGEN_LOCAL_CACHE_DIR: Local worker cache directory (default /_local_cache/datagen).
• DATAGEN_ACTIVE_POLL_SECONDS: UI poll interval while tasks are active (default 10).
• DATAGEN_IDLE_POLL_SECONDS: UI poll interval when idle (default 30).
• DATAGEN_REGISTER_RATE_WINDOW_SECONDS: Registration rate-limit window seconds (default 900).
• DATAGEN_REGISTER_MAX_PER_IP: Max registration attempts per IP per window (default 10).
• DATAGEN_REGISTER_MAX_PER_EMAIL: Max registration attempts per email per window (default 3).
• DATAGEN_REGISTER_RESEND_WINDOW_SECONDS: Resend rate-limit window seconds (default 900).
• DATAGEN_REGISTER_RESEND_MAX_PER_IP: Max resend attempts per IP per window (default 5).
• DATAGEN_REGISTER_RESEND_MAX_PER_EMAIL: Max resend attempts per email per window (default 2).
• DATAGEN_REGISTER_FAILURE_TTL_SECONDS: TTL for failed-email resend records (default 86400).

Current API Surface

• POST /api/auth/login: Authenticate via cstore-auth-ts; sets HttpOnly session cookie.
• POST /api/auth/logout: Clears the session.
• GET /api/auth/me: Returns current session (401 when missing/invalid).
• POST /api/auth/register: Creates account and emails generated credentials.
• POST /api/auth/register/resend: Re-sends credentials for recent failed delivery attempts.
• GET /api/metrics: Auth-protected metrics from persisted CStore counters.

Project layout (high level)

app/
(auth)/login/page.tsx
(auth)/register/page.tsx
(app)/page.tsx
api/
auth/login|logout|me|register|register/resend
metrics/route.ts
lib/
auth/
datagen/
ratio1/

Citations

@software{smartclover_datagen_2026,
author = {{SmartClover SRL}},
title = {DataGems: Synthetic Dataset Generation Platform},
year = {2026},
version = {0.5.0},
url = {https://github.com/SmartCloverAI/DataGems},
organization = {SmartClover SRL},
note = {Accessed 2026-02-17}
}

@article{Nyanchoka_2022,
title = {Understanding facilitators and barriers to follow-up after abnormal cervical cancer screening examination among women living in remote areas of Romania: a qualitative study protocol},
author = {Nyanchoka, Linda and Damian, Andreea and Nyg{\aa}rd, Mari},
journal = {BMJ Open},
volume = {12},
number = {2},
pages = {e053954},
year = {2022},
month = feb,
doi = {10.1136/bmjopen-2021-053954},
url = {https://doi.org/10.1136/bmjopen-2021-053954}
}

@misc{smartclover_cerviguard_pilot_2026,
title = {SmartClover CerviGuard Pilot},
author = {Andreea D and Cristian Bleotiu and Vitalii Toderian and Florian Nicula},
year = {2026},
howpublished = {\url{https://github.com/SmartCloverAI/CerviGuard}},
note = {Pilot web console for cervical image analysis and case management; citation metadata as published in the CerviGuard repository README}
}

References

• DataGems production app: https://datagems.app
• SmartClover official site: https://smartclover.ro/
• SmartClover About: https://smartclover.ro/about
• SmartClover Products & More: https://smartclover.ro/products
• CerviGuard public workspace: https://cerviguard.link