Supercharge Your Obsidian Vault with Semantic Search

If you've been using Obsidian for a while, you know the feeling: your vault grows into hundreds (or thousands) of notes, and keyword search starts letting you down. You remember writing something about "that productivity idea" but can't recall the exact words you used. You search for "focus" and miss the note where you wrote about "deep work." You search for "anxiety" and miss the one where you described "that overwhelming feeling before a big presentation."

Obsidian's built-in search is fast and powerful but it's literal. It finds what you typed, not what you meant.

The obvious workaround feeding your notes to an AI assistant helps, but at a cost. Loading your vault into a language model's context window means paying for tokens every time you search. With a large vault, that gets expensive fast. And you're sending your private notes to a third-party server in the process.

What is ThinkableSpace?

ThinkableSpace brings semantic search to your local documents. It runs entirely on your machine. It uses a locally-run embedding model (Google's EmbeddingGemma-300m via llama.cpp) to understand the meaning of your documents, and HNSW vector search to find relevant content in milliseconds.

It supports Markdown natively, which means your Obsidian vault is a first-class citizen.

Setting It Up

Getting your Obsidian vault into ThinkableSpace takes about two minutes:

1. Open ThinkableSpace and go to Settings > Sync Folder
2. Add your Obsidian vault directory (e.g. ~/Documents/MyVault)
3. ThinkableSpace will scan and index all your .md files in the background
4. Once indexing completes, open the search panel and start searching using natural-language queries

That's it. No plugins needed on the Obsidian side, no exports, no syncing to a third party.

What Changes When You Have Semantic Search

Here's what becomes possible:

Ask questions instead of guessing keywords. Instead of searching meeting notes project alpha, you can search for "what did I decide about the alpha project timeline?" and surface the relevant notes even when the wording is completely different.

Rediscover forgotten notes. You wrote something insightful six months ago and completely forgot about it. Semantic search connects the idea you're thinking about now to the note you wrote then, even if the vocabulary doesn't overlap.

Find notes by concept, not by title. Your note titled "Monday – random thoughts" might contain your best idea of the year. Keyword search buries it. Semantic search finds it when you're exploring that concept weeks later.

Surface patterns across your vault. Search for "things I've written about decision-making under uncertainty" and get a ranked list of relevant passages across dozens of notes, your own thinking surfaced and ranked by semantic relevance.

Keyword search is still useful for exact matches, filenames, code snippets, and tags, semantic search complements it by handling conceptual recall.

A Few Things to Know

Obsidian-specific syntax is included as-is. Wiki links like [[Note Title]], YAML frontmatter, and #tags are part of the indexed text. This generally doesn't hurt results, the embedding model handles noisy formatting well but it's good to know.

Attachments are handled by type. If your vault has embedded PDFs, ThinkableSpace will index those too. Images are skipped. Audio/video files are skipped.

The .obsidian/ config folder inside your vault holds JSON configuration files that ThinkableSpace will skip (unsupported format), so no noise from Obsidian's internal settings.

Changes are indexed automatically. ThinkableSpace watches your vault folder. New notes, edits, and deletions are picked up within a second or two, no manual re-indexing required.

Everything stays local. Your notes never leave your machine. The embedding model runs on-device, the vector index lives on disk, and there's no telemetry on your content.

Who This Is For

This setup is particularly useful if you:

• Keep a personal knowledge base, Zettelkasten, or second brain in Obsidian
• Write daily notes and want to surface relevant past entries
• Research topics over long periods and want to connect ideas across time
• Have a large vault where Obsidian's native search feels like searching in the dark

The Bigger Picture

Obsidian is excellent at organizing your thinking. ThinkableSpace makes that thinking retrievable, not just the notes you remembered to tag correctly, but all of it.

Your vault becomes less of an archive and more of an extension of memory: something you can query the way you query your own mind, with fuzzy, meaning-driven recall rather than brittle keyword matching.

If you've been looking for a reason to be more prolific in Obsidian, to write freely without obsessing over tags and structure, semantic search is it. Write naturally. Find it later.