Δ emetra

Generator

Realistic CSV generation by column schema — built for pipelines that later need anomaly checks, not just pretty demo rows.

• Schema-first designer Define columns, types, blank rates, and constraints — export CSV, JSON, or SQL ready for loaders.
• Detector-ready fixtures Templates tuned for IoT, finance, and logistics so generated files exercise Detector the way production data would.
• Controlled chaos Inject reproducible outliers, drift, and mixed types with a seed — unique vs generic mock tools that only aim for “looks real.”
• API & CLI generation Pull rows programmatically for CI and load tests; keep schemas versioned with your repo.

Detector

Upload a CSV — Demetra infers column types, runs statistical and ML detectors, and explains what looks wrong.

• Understands your columns Numbers, dates, emails, IPs, statuses, SKUs, coordinates — types from values, not just headers.
• Multi-angle detection PyOD / scikit-learn checks, group patterns, and clustering so unusual rows stand out clearly.
• Explained findings Each anomaly comes with a short rationale so you can fix or ignore without reading every raw row.
• Data lab workflow Upload → job → column profiles → anomalies — from file to insight in a few steps.

Collector

An open API that collects and structures headlines from news sources — ready for research, alerts, and downstream models.

• Multi-source ingest Pull fresh items from configured news sites on a schedule, with idempotent upserts by URL.
• Open developer API Query recent articles, filter by topic or time window, and wire results into your own apps.
• Structured payloads Consistent fields for title, source, URL, timestamps, and tags — not scraped HTML dumps.
• Feeds Forecast The same news stream powers trading analytics when you need market context, not just a feed reader.

Forecast

Trading analytics that combine news, trends, and prices into next-session research signals for a curated ticker universe.

• Signal fusion News, Reddit, Google Trends, and US equity prices analyzed into shared dimensional fields.
• Similar situations Embeddings and vector search surface historical days that look like today — before the next open.
• Next-session outlook Research forecasts for NYSE/NASDAQ tickers you opt into — not a black-box trading bot.
• Transparent pipeline Collect → analyze → embed → graph — inspect each stage instead of trusting a single score.

Research signal only — not financial advice.