Vectile

A fully local, privacy-preserving RAG (Retrieval Augmented Generation) system for Windows, macOS, and Linux. It indexes personal knowledge from several sources into a single SQLite database with hybrid vector + full-text search, then lets you find things by meaning, not just by exact words, from a fast keyboard-first desktop app. Everything runs on your machine. No server, no cloud, no network calls.

Inspired by Sebastian Hutter’s local-rag. No Ollama, no API keys. The embedding model runs in-process from a .gguf file: import one of your own, or download one from the built-in catalog. Either way, nothing leaves your machine.

Features

• Fully local and private. Searches run on your machine against an in-process embedding model. No server, no cloud, no telemetry, no account.
• Hybrid search. Vector and full-text results are fused with Reciprocal Rank Fusion, so a query can find a note that never uses your exact words.
• Index what you keep. Obsidian vaults, project folders of documents (Markdown, PDF, DOCX, HTML, TXT, CSV, JSON, YAML, XML, SQL, shell, XLSX, PPTX, Jupyter notebooks, EPUB), Calibre libraries, and code repositories including their commit history.
• Built-in model manager. Import your own .gguf, pick the active model, or download one from the curated catalog in Settings with a live progress bar.
• Keyboard-first desktop UI. Jump to search from anywhere with ⌘K / Ctrl K, and move between Search, Library, Browse, Index, and Settings from the sidebar.
• Manage your library. Expand a collection to its files, drill into individual chunks, and delete stale sources, selected chunks, or a whole library in place.
• AI assistant access (MCP). Serve search and collection tools to Claude Desktop or any MCP client over a local server, with index and prune tools available behind an Allow write tools toggle.
• A little company. Vexter, the pixel dinosaur, pokes up in the sidebar while you search, index, or come up empty. Settings → Vexter controls each moment.

Download

Click your platform to download the latest version:

Windows - ⚠️ SmartScreen will block it · how to fix
Windows portable
Linux AppImage
Linux deb
macOS universal

Windows 10/11 · Linux (AppImage + deb) · macOS (universal arm64 + amd64) · app is not code-signed, see first-run notes below

First-Run Notes

vectile is not code-signed, so your OS may warn you on first launch. The app is safe and fully open source — you can read every line of code here.

Windows - SmartScreen

1. Click More info
2. Click Run anyway

Or right-click the .exe -> Properties -> check Unblock -> Apply.

macOS - Gatekeeper

Right-click the app -> Open (once), or run xattr -d com.apple.quarantine /path/to/vectile.app.

Linux

The .deb pulls GTK4/WebKitGTK 6.0 + libgomp1 automatically; the AppImage needs chmod +x before running.

Screenshots

Screenshots show sample data.

Supported sources

Source	Collection Type	What Gets Indexed
Obsidian	system	Vault files: .md notes with frontmatter, tags, and wikilinks
Project folders	project	Any folder of documents, each file parsed by its extension (.md, .pdf, .docx, .html, .txt, .csv, .json, .yaml, .xml, .sql, .sh, .xlsx, .pptx, .ipynb, .epub)
Code repositories	code	Git repos: tree-sitter splits each function and class into its own chunk (cAST split-then-merge); commit history is indexed as its own source
Calibre	system	Ebook metadata + content: title, author, tags, series, publisher, description, and EPUB/PDF text

Installation

From source

Prerequisites:

• Go 1.26+
• Node.js + npm
• the wails3 CLI
• Windows only: MinGW-w64 on PATH, with LIBRARY_PATH and C_INCLUDE_PATH pointing at third_party/llama-go (the Windows build task sets these)

From the project root:

task dev          # run in development mode
task build        # build the binary to bin/
task package      # package an installer for the current OS

Installing the model

Drop a .gguf file into the models/ folder inside the app data directory, import one from Settings (a native file dialog copies it into models/), or grab one from the curated catalog right in the app — Settings → Model → Get a model downloads an embedding model with a click, with a live progress bar. The default model is bge-m3.

Quick start

On a fresh install, a short tour walks you through it: add a folder, index it, then search.

1. Launch vectile.
2. Download or import an embedding model in Settings, or drop a .gguf into models/.
3. Add sources in Settings: Obsidian vaults, project folders, code repositories, Calibre libraries.
4. Open the Index view and index a collection, or everything at once. Unchanged files are skipped, so re-indexing is fast.
5. Press ⌘K / Ctrl K and search.

GUI

Five views, keyboard-first:

• Search (home): a large search bar, a filter row, and results as cards with title, snippet, rank, collection, and source path. Expand a card to read the whole passage, open the file, or reveal it in the file manager. A small toggle switches each result between its rank (#1) and the blended score (%). Jump in from anywhere with ⌘K / Ctrl K.
• Library: every collection with its file and chunk counts and the last time it was indexed; expand one to list its files, and remove a source or its documents in place.
• Browse: a file tree of collections, files, and chunks, with a preview pane. Select chunks to delete them, or remove a whole library.
• Index: run "Index new" (only changed files) or "Re-index all" (re-embed everything) per collection, or index all collections at once, with live progress.
• Settings: sources, model (download an embedding model from the curated catalog, or import your own), chunking, search defaults, auto-reindex, start-on-login, Vexter (the sidebar mascot), and a Connect section that runs a local MCP server for AI assistants.

The sidebar shows the model state: idle, loaded, or failed. If the model file is missing or corrupt, vector search falls back to full-text search, so exact-word matches still work.

A little pixel dinosaur called Vexter lives in the sidebar and pokes up while your library works — while a query runs, while it indexes, and when a search comes up empty. It's purely decorative, and Settings → Vexter lets you turn each of those moments on or off independently.

Files you delete get pruned automatically, so results don't go stale. Auto-reindex, if enabled, re-indexes everything on a timer. Start-on-login launches the app with your session. The status strip shows when the library was last indexed; when auto-reindex is off and that date is more than a day old, Search quietly suggests a re-index.

AI assistants (MCP)

Settings → Connect runs a local MCP (Model Context Protocol) server on 127.0.0.1:31123. It exposes search and collection tools, plus index and prune tools that stay off until you enable Allow write tools in Settings. The server binds to loopback only, so nothing leaves the machine.

Point Claude Desktop, Claude Code, or any MCP client at http://127.0.0.1:31123/sse to search your library from the assistant. The Settings section shows the live server status, the tools it serves, and per-client setup directions.

How search works

A query runs two searches at once.

• Full-text search matches the exact words against an FTS5 index. Fast, precise, literal.
• Vector search embeds the query and finds stored vectors that point the same way. That's how a query like "how do we ship changes safely" can match a note about blue-green deploys that never uses those words.

The vector path is two-stage: a cheap binary-quantized index finds a pool of candidates, then the exact float vectors are fetched and reranked by distance. The two result lists are merged with Reciprocal Rank Fusion, which blends ranks rather than scores.

Filters narrow results: collection, source type, path substring, sender or author, and date range. Top-k controls how many results come back.

Configuration

Config file: <os.UserConfigDir()>/vectile/config.json

Key	Default	Description
embedding_model	bge-m3	Embedding model name
active_model	(default model path)	Path to the active .gguf model
embedding_batch_size	32	Chunks per embedding call
chunk_size_tokens	500	Chunk size in whitespace-separated words
chunk_overlap_tokens	50	Overlap between chunks
obsidian_vaults	[]	Paths to Obsidian vaults
obsidian_exclude_folders	[]	Folders to skip in vaults
calibre_libraries	[]	Paths to Calibre libraries
repositories	{}	Map of collection name to repo or directory paths; directories are scanned recursively for git repos
projects	{}	Map of collection name to document paths
disabled_collections	[]	Collection names to skip during indexing
skip_cloud_placeholders	true	Skip cloud-only placeholder files (OneDrive, iCloud, Google, Synology) instead of downloading them
git_history_in_months	6	How far back to index commit history
git_commit_subject_blacklist	[]	Skip commits whose subject starts with any of these strings
search_defaults.top_k	10	Default number of search results
search_defaults.rrf_k	60	Reciprocal Rank Fusion parameter
search_defaults.vector_weight	0.7	Weight for vector similarity
search_defaults.fts_weight	0.3	Weight for full-text search
gui.auto_reindex	false	Enable periodic re-indexing
gui.auto_reindex_interval_minutes	60	Minutes between auto-reindex runs
gui.start_on_login	false	Launch at login
gui.mascot.show_searching	true	Show Vexter in the sidebar while a query runs
gui.mascot.show_indexing	true	Show Vexter while a library rebuilds
gui.mascot.show_nothing	true	Show Vexter when a search comes up empty
mcp.enabled	false	Serve MCP tools to local AI assistants on launch
mcp.port	31123	Port the MCP server listens on (127.0.0.1 only)
mcp.allow_write	false	Let AI assistants call the index and prune tools

Tech stack

Component	Choice	Notes
Language	Go 1.26+	Wails v3 desktop app
UI	SolidJS + TypeScript + Vite	Tailwind CSS v4
Database	SQLite (modernc.org/sqlite) + sqlite-vec + FTS5	Pure Go, no cgo; single file
Embeddings	llama.go (llama.cpp)	In-process .gguf; bge-m3 by default; no Ollama
Code parsing	go-tree-sitter	Structural splitting (functions, classes, methods) with the cAST split-then-merge strategy
PDF	go-pdfium (WASM/Wazero)	No cgo needed
DOCX	archive/zip + encoding/xml	Word document extraction (.docx, .dotx)
XLSX	excelize	Spreadsheets, one ## Sheet section per worksheet
PPTX	archive/zip + encoding/xml	Slides, one ## Slide section per slide

Building and developing

task dev        # run in development mode
task build      # build the binary to bin/
task package    # package an installer for the current OS

Tests: go test ./backend/.... Model-dependent tests skip when the model is not in models/.

vectile links llama.cpp in-process through the vendored third_party/llama-go, whose static archives are committed per-OS/per-arch under third_party/llama-go/{windows,linux,darwin}/<arch>. A build only needs a C/C++ compiler on PATH; see docs/BUILD-AND-PACKAGING.md for the full cross-platform build, packaging and release notes.

• Windows (amd64): needs MinGW-w64 on PATH, with LIBRARY_PATH/C_INCLUDE_PATH pointing at third_party/llama-go (the Windows build task sets these). The built exe needs five MinGW runtime DLLs beside it (libgcc_s_seh-1.dll, libgomp-1.dll, libstdc++-6.dll, libwinpthread-1.dll, libdl.dll). Missing libdl.dll causes a silent 0xC0000135 exit at launch.
• Linux (amd64): needs gcc/g++, pkg-config, libgtk-4-dev, libwebkitgtk-6.0-dev and libayatana-appindicator3-dev. The binary depends on libgomp.so.1 at runtime (bundled in the AppImage, Depends: libgomp1 in the .deb). task linux:package yields AppImage + .deb (+ .rpm/AUR).
• macOS (universal): needs Xcode Command Line Tools. task darwin:build:universal builds arm64 + amd64 and lipos them together; task darwin:package:dmg wraps the .app in a DMG. The Metal/Accelerate frameworks are OS-provided, so nothing extra ships.

Architecture

main.go                     app startup, window, services, auto-reindex loop
backend/appdata             the data directory and the model path
backend/config              config.json load, save, defaults
backend/db                  SQLite schema and helpers (modernc + vec0 + FTS5)
backend/embeddings          the llama.go embedder (bge-m3)
backend/chunker             word-window and markdown chunking
backend/parser              file parsers: md, docx, html, epub, pdf, xlsx, pptx, ipynb, xml, sql, shell, csv/json, calibre, code
backend/search              hybrid search: vector + FTS + RRF
backend/indexer             obsidian, project, git, calibre indexers; prune
backend/services            Wails services the UI calls
backend/startup             launch-at-login per OS
third_party/llama-go        vendored llama.cpp bindings
frontend/src/lib/api.ts     the only place the UI touches the bindings

License

MIT License, copyright (c) 2026 d3uceY.