dsh-data-insight

Data analysis toolkit for DeepSeek Harness. Drop a CSV and get instant profiling, anomaly detection, and queryable insights — all without leaving your agent session.

Features

• profile_data — Column types, missing values, statistics, distributions
• detect_anomalies — IQR and z-score outlier detection on numeric columns
• summarize_data — Structured overview with quality issues and next-step suggestions
• query_data — Filter, group, and aggregate without writing code

Install

dsh plugin add github:ClaireXi99/dsh-data-insight

Usage

Once installed, the tools are available to the agent automatically.

> Profile the sales data in ./data/q3_sales.csv

> Detect anomalies in the "revenue" column of report.csv

> Summarize the dataset at ./logs/user_events.json

> Query users.csv where age > 30, group by city, show mean salary

Supported Formats

• CSV (comma-separated)
• TSV (tab-separated)
• JSON (array of objects, or {data: [...]} / {results: [...]})

Development

git clone https://github.com/ClaireXi99/dsh-data-insight.git
cd dsh-data-insight
npm install
npm run build
npm test

Local dev with DSH

dsh web --patch ./dev/cordis.yml

License

MIT