Home

SenseTree Wiki

SenseTree is a local-first semantic overlay for your Windows file system: search files by meaning, chat with a folder, and let an AI propose (never silently execute) reorganizations — with the AI models you choose, running entirely on your machine.

This wiki is the complete documentation. Use the sidebar, or start here:

New here?

• Installation — download and install the app.
• Getting Started — index your first folder and run your first search.
• FAQ — quick answers (updates, privacy, data location, MSI vs EXE…).

Using SenseTree

• Configuration — every setting, and where settings.json lives.
• Models & Providers — the model-agnostic design, local vs. Ollama/LM Studio, the live model catalog, and hardware recommendations.
• Semantic Search — how meaning-based search and scoping work.
• AI Chat & Dry-Run Actions — chat with a folder and review Before→After plans safely.
• Gardener — audit a directory for duplicates, empty folders, and clutter.
• Prompts — tune the AI's behavior without recompiling.

Under the hood

• Indexing Pipeline — crawler, watchdog, worker, and the block/recursive classifier.
• Architecture — module map, data stores, and the IPC surface.
• Building from Source — dev setup and the release process.
• Troubleshooting — fixes for the issues you're most likely to hit.

---

What makes it different

Real files, not a vault	SenseTree overlays your actual folders. It never hides or relocates files behind your back.
Meaning, not keywords	Vector embeddings let you search by concept and browse a meaning tree.
You propose, I plan	Structural changes are always shown as an interactive Before → After diff; apply is transactional with rollback.
Bring your own models	Embedding, reasoning, and vision are three independent slots — local engine or any OpenAI-compatible server.
Private by construction	Fully offline. Embeddings are computed internally; LLM calls go only to the endpoints you configure. No telemetry.

Current version: 1.1.0. See the Releases page for installers and changelog.