Getting Started

Getting Started

This walks you from a fresh install to your first semantic search and Dry-Run action.

The layout

SenseTree is a three-panel desktop app:

• Sidebar (left) — your indexed roots, health indicators, and the Settings button.
• Explorer (center) — breadcrumb + file list of the current folder, with semantic tags and per-file indexing status. Switches to a search-results view when you search, and offers a meaning tree view.
• Chat (right) — talk to the current folder; when the assistant proposes file operations, they render as a Before → After action card.

Step 1 — Add a folder to index

In the sidebar, click add a folder and pick a root (e.g. Documents). SenseTree starts:

• a crawler that walks the tree (the "past"), and
• a watchdog that reacts to live changes (the "present").

You can add several roots and remove any of them later. See Indexing Pipeline.

Step 2 — Choose your models

Open Settings (gear icon). There are three independent slots:

Slot	What it does	Default
Embedding	Turns text/captions into vectors for search.	Built-in local multilingual-e5-small (384-d, multilingual).
Reasoning	The LLM for chat, reorganization plans, and folder descriptions.	llama3.1:8b via http://localhost:11434/v1 (Ollama).
Vision	Multimodal model for image captions and PDF OCR.	moondream (Ollama), disabled by default.

Embedding works out-of-the-box with no extra software. For reasoning/vision, point the base URL at your Ollama or LM Studio server, or browse the live model catalog to download one. Use Test connection to verify each endpoint; the header shows overall AI health.

Changing the embedding model (or its dimensions) triggers a full re-index, because vectors of different models aren't comparable. See Configuration.

Step 3 — Let it index

Watch progress in the header (queued / completed counts). You can pause/resume indexing at any time — it frees the model and resumes where it left off. Each file gets one of three kinds of "sense":

• Textual — real content extracted from PDFs, Office docs, text, and code.
• Visual — a caption from the vision model (if enabled), else context.
• Contextual — for opaque/huge files: filename + parent folder + type.

Step 4 — Search by meaning

Type a concept in the search bar — "insurance renewal", "trip to Korea", "budget spreadsheet". Results are ranked by semantic similarity with snippets, and you can scope the search to the current folder. See Semantic Search.

Step 5 — Chat and (safely) reorganize

Open the chat panel and ask a question about the folder, or an action like "group these invoices by year". If it's a question, you get a normal answer citing files. If it's an action, you get a Dry-Run plan: a Before → After list of moves/renames you can Approve or Discard. Nothing is written to disk until you approve, and apply rolls back cleanly on any failure. See AI Chat & Dry-Run Actions.

Where's my data?

Everything lives under %APPDATA%\com.virgi.sensetree:

• settings.json — your configuration,
• a SQLite DB — file metadata, the indexing queue, and extracted summaries,
• a LanceDB directory — the vectors.

Nothing is uploaded anywhere. See the FAQ.