Hybrid Search

Fast hybrid search for Obsidian — combines BM25 full-text and semantic (vector) search over your vault.

The plugin is a thin UI layer that communicates with the obsidian-hybrid-search CLI, which runs as a background process and handles indexing and search.

Desktop only.

Requirements

Install the CLI globally via npm:

npm install -g obsidian-hybrid-search

The CLI must be available in your PATH (or you can set a custom path in plugin settings).

Installation

Via Obsidian Community Plugins

Open the Hybrid Search page and click Add to Obsidian.

Via BRAT

1. Install the BRAT plugin.
2. In BRAT settings, click Add Beta Plugin and enter this repository's URL.
   • https://github.com/flowing-abyss/obsidian-hybrid-search-plugin
3. Enable Hybrid Search in Obsidian's plugin list.

Usage

Open the search modal with:

• Ribbon icon — click the search icon in the left sidebar.
• Command palette — run Hybrid search: Open search.

Type to search. Results appear as you type with a relevance score:

📷 Hybrid search results with relevance scores

Score	Color	Meaning
>0.8	Green	High relevance
0.5–0.8	Orange	Medium relevance
<0.5	Gray	Low relevance

Empty query & similar notes

Leave the query empty and the plugin shows contextually relevant results:

• If a note is open, it surfaces semantically similar notes.
• If no note is open, it shows recently opened files.

📷 Semantically similar notes

Workbench

The plugin includes a link-discovery workbench that analyzes your vault's knowledge graph to surface the most actionable connections for the current note. It runs several graph algorithms side by side:

• Best — top candidates ranked by a composite of structural and semantic signals.
• Missing Links — notes that are strongly related but not yet linked.
• Bridges — notes that connect otherwise distant parts of your graph.
• Similar — semantically close notes.
• Links — existing outgoing and incoming links.
• Diagnostics — structural insights about the note's position in the graph.

Each candidate exposes interpretable scores — cosine, Adamic-Adar, common neighbors, co-citation count, semantic distance, and more — so you can decide whether a suggested link makes sense. You can add links, inspect backlinks, or compare notes directly from the list, while a live local graph visualizes the neighborhood around the current file.

📷 Link-discovery workbench

Query Syntax

These two queries are equivalent:

hybrid: zettelkasten tag:project limit:20 @rerank
zettelkasten #project @limit:20 @rerank @hybrid

Inline	Postfix	Description
(plain text)	—	Hybrid search (default)
hybrid:	@hybrid / @hyb	Hybrid mode (BM25 + semantic)
semantic: / sem:	@semantic / @sem	Semantic (vector) only
fulltext: / full:	@full	Full-text (BM25) only
title:	@title	Fuzzy title match
tag:	#tag	Filter by tag (include)
tag:-	-#tag	Filter by tag (exclude)
folder:	—	Limit to a folder
limit:N	@limit:N / @lim:N	Override result count
threshold:N	@threshold:N / @th:N	Minimum score threshold
—	@rerank	Re-rank with cross-encoder
key:value	—	Filter by frontmatter field

Filters can be combined freely.

Hotkeys

Hotkey	Action
Mod+J / Mod+K	Next / previous result
Mod+P	Toggle preview panel
Mod+G	Toggle local graph panel
Mod+O	Open selected in new tab
Mod+Shift+O	Open all results in new tabs
Alt+Enter	Insert wiki link to selected note
Alt+Shift+Enter	Insert wiki links to all results

Mod = Ctrl (Windows/Linux) or Cmd (macOS). Alt = Option on macOS.

License

MIT