Pre-launch First public release in progress. Watch on GitHub →
MyDataSights

Pre-release · v0.1 in development

A private, offline-first,
portable knowledge workspace.

Drop any file. Find anything. Run AI locally.
No account. No cloud. No telemetry. After your chosen models are installed, no internet is required at runtime.

Watch on GitHub See how it works

Try-in-browser and desktop installers ship at v0.1 launch.

app.mydatasights.app

Inputs

📄 air-quality.csv

📊 electrical-3phase.csv

📚 books-library.csv

🎲 boardgames.csv

Insights

⚡ PM2.5 breach detected — bin 14:00–15:00

📈 RADON ~ NO2 correlate (r=0.71)

🔍 Profile: air-quality (confidence 0.94)

🧮 Drift on column “temp”: stable → drifted

Models

⓿ Phi-3-mini-Q4

⓿ MiniLM-L6

⓿ Whisper-tiny

⓿ all on-device

Mockup of the v0.1 workspace — interactive preview at launch.

What it does

Ingest anything

Text, CSV, JSON, Markdown, code, images, audio, video — eight parsers and growing. Drop a folder; it understands the contents.

Search by meaning

Full-text plus embeddings plus tags plus facets — one search box, four index layers, ranked together.

Visualise

Time-series, heatmaps, latent projections, mind-maps, timelines. The right view for the data, picked automatically.

Run AI locally

Sentiment, NER, summarisation, OCR, chat. Models live on your device. Nothing leaves it.

Extend

Plugins are folders with a manifest. Drop one in — parser, viewer, analyser, view. No rebuild required.

Private by design

No server, no analytics, no account. Even crash reports stay in your browser until you choose to export them.

How it works

  1. Drag a folder in. Parsers pick up CSVs, Markdown, JSON, code, images, audio, video.
  2. It indexes and enriches. Full-text index, language detection, keywords, statistics, domain profile.
  3. You ask, chart, analyse. Search by meaning, visualise, run local AI — all on your machine.

Frequently asked

When does this launch?

v0.1 is in active development — track progress in the project status doc on GitHub. The fastest way to be notified at launch is to Watch → Releases only on the repository. No mailing list, no signup form: we hold no email addresses we don't have to.

Does anything leave my computer?

Only AI model files, and only when you choose to download one. After that, everything — your files, your searches, your analyses — stays on-device. See privacy for the exact list.

Do you collect analytics?

No. There is no telemetry, no cookies, no tracking pixels. The only server-side data is your IP in Cloudflare's access logs while you visit this page, retained briefly for abuse prevention.

What's the difference between the browser and desktop versions?

The browser version runs the same app, but is limited by browser storage quotas and cannot reach files outside what you drag in. The desktop version (Tauri) can read your filesystem, has larger storage, and supports bigger AI models like the 2 GB Phi-3 chat model.

Is it really free?

Yes. The open-source core is Apache-2.0 — no account required, no SaaS upsell. An optional paid Pro edition with extra plugins (encrypted workspace, advanced analyzers) is planned for v1.0; the free core will always remain fully functional on its own. Sponsoring on GitHub is welcome but optional.

Can I use it commercially?

Yes. The Apache-2.0 license permits commercial use; see the license. Outputs are advisory and you should verify before relying on them for any business decision.

How accurate are the AI outputs?

The app surfaces patterns and statistical estimates, not authoritative truth. Models can be wrong — sometimes confidently — on small samples, ambiguous text, or data outside their training domain. Do not rely on outputs for medical, legal, financial, or safety-critical decisions.

How do I report a security issue?

Email [email protected]. Please don't open public issues for vulnerabilities.

Star it now. Use it at launch.

No email signup. Just a GitHub watch for release notifications.

Watch on GitHub → View source