Developer JASWANT-KUMAR-NAYAK has released Hirune, a desktop application designed to reconstruct your workday using activity logs rather than invasive screenshots. The tool captures active application names and window titles, organizes them into sessions, and uses AI to generate structured daily reports. Unlike many productivity trackers that rely on visual surveillance, Hirune focuses on metadata and local storage to maintain user privacy while providing actionable insights.

Privacy-First Architecture

Hirune operates on a strict 'no capture' policy for sensitive data types. The application explicitly does not record screenshots, keystrokes, clipboard contents, audio, or URLs. All activity records are stored locally in a SQLite database, ensuring that raw data never leaves the user's machine unless they choose to export it. This design choice addresses the growing developer skepticism toward 'bossware' tools that feel more like surveillance than support, positioning Hirune as a self-reflective utility rather than an employer's monitoring tool.

Open-Weight AI Integration

The reporting engine relies on Gemma 4 (31B), an open-weight model accessed via the Ollama Cloud API. This choice allows Hirune to generate human-readable summaries from raw activity logs without locking the user into a proprietary ecosystem. The system is built with Python for backend logic and uses HTML, CSS, and JavaScript for the dashboard. By decoupling the AI inference layer from the core tracking logic, the developer has made it easier to swap out models or inference providers in the future, supporting a modular approach to AI integration in dev tools.

Verifiable Evidence Over Black Boxes

A standout feature of Hirune is its emphasis on evidence-backed claims. Every AI-generated statement in the daily report includes references to the specific activity records that support it. Users can review these sessions, edit labels, and add personal notes, creating a feedback loop where the human verifies the machine's interpretation. This 'human-in-the-loop' verification process prevents the tool from becoming an opaque productivity score, instead offering a collaborative reconstruction of the dayโ€™s work. The data can be exported to CSV, giving users full ownership and portability of their activity history.

Key Takeaways

  • Hirune uses Gemma 4 via Ollama Cloud for AI summaries, avoiding proprietary model lock-in.
  • The tool explicitly avoids capturing screenshots, keystrokes, or audio to prioritize privacy.
  • AI-generated reports include direct references to underlying activity records for verification.
  • All data is stored locally in SQLite, with CSV export options for user-owned records.

The Bottom Line

Hirune is a refreshing take on productivity tooling that respects developer autonomy. By prioritizing local storage and verifiable AI claims over invasive surveillance, it offers a practical alternative to the current wave of enterprise monitoring software. The use of open-weight models like Gemma 4 signals a maturing trend in dev tools: moving away from black-box SaaS dependencies toward transparent, user-owned AI workflows. For developers who want to understand their time usage without feeling watched, this is a solid, privacy-conscious foundation to build on.