Most analysis servers stop at producing output; this one delivers it, which means you are handing an assistant a working sending credential on your address. Use a dedicated account and an app password, and be deliberate about what goes into a draft — the project's own notes make the same point. The analysis half stands on its own if you would rather not wire up email at all.
Three steps of a reporting workflow in one server: statistical analysis of a data file, generation of written research content, and delivery by email — with the subject line itself written by the model.
- `analyze-data` takes an Excel or CSV file, detects and parses types, runs statistics over numeric columns including correlation and outlier detection, and writes an HTML report with interactive Chart.js plots — with a basic and a detailed mode and a configurable output directory
- `generate-thinking` produces research papers, technical documentation, data-analysis summaries and business-intelligence reports from a prompt, saving each response with a timestamp
- `send-email` delivers the result over SMTP with plain text and HTML bodies, image attachments that can display inline, and a subject line generated from a prompt rather than typed
- Output lands in a predictable structure — analysis reports and plots, generated content and email logs each in their own folder
Node.js v16 or higher and TypeScript. Clone, install and build, then point the client at the compiled entry point; published as gemini-email-subject-generator, version 1.0.0. Three credentials are needed: GEMINI_API_KEY for the model, and NODEMAILER_EMAIL with NODEMAILER_PASSWORD for sending — for Gmail that has to be an app password rather than the account password. Supported input formats are .xlsx, .xls and .csv, in UTF-8.
Build from source — clone the repository and build it, then point your client at the binary