BYO AI: a desktop work log that refuses to phone home

What it does

WorkShadow is a Tauri-based desktop app that acts as a daily work journal. It stores rich-text logs locally in SQLite, indexes them for keyword and semantic search, and offers a “workbench” where you can summarize multiple entries into reports or ask questions across your own history. If you want AI features—like vector search, log summarization, or image captioning—you bring your own model endpoints rather than relying on a vendor’s cloud.

The interesting bit

The project draws a hard line between storage and intelligence: your data never leaves the machine unless you explicitly wire up an LLM or embedding model. Even the smart-completion feature (in the installer edition) runs inference locally and trains on your past logs without uploading them, which turns the app into a quiet, personal pattern-matching engine.

Key highlights

• Data stays local by default; AI only activates if you configure your own models.
• Semantic search via LanceDB finds conceptually related logs even when keywords differ.
• The Workbench can draft daily, weekly, or project reports from selected log entries.
• Logs, settings, and “memories” (cross-project context) pack into .ws files for backup or migration.
• The installer edition adds local model inference, unlimited summarization, and smart completion; the dev edition is feature-gated.

Caveats

• AI features—including semantic search, summarization, and Q&A—require manual model configuration; there is no built-in managed AI backend.
• Several quality-of-life features (local model inference, smart completion, code syntax highlighting) are reserved for the installer edition and absent from the open-source dev build.

Verdict

Worth a look if you want a private, offline-capable work diary with optional AI augmentation. Skip it if you need a collaborative or fully managed SaaS note-taking experience.