SnipsetDocs
FeaturesPricingBlogDownload
Documentation

Productivity Tools

OmniSearch

Use AI semantic search to find snippets by meaning, not just exact words.

Regular search looks for exact text matches. OmniSearch understands meaning. If you search for "follow up with a client", OmniSearch can find your "Check-in email" snippet—even if the exact words "follow up" don't appear anywhere in the snippet's content.

What Are Embeddings?

OmniSearch works by converting every snippet into a "vector embedding"—a long list of numbers that mathematically represents the meaning of the text.

  • When you perform a search, your search query is also converted into an embedding.
  • Snipset then compares the numbers mathematically to find snippets with similar meanings.
  • This translation is done locally on your computer by the embedding model via Ollama (for example hf.co/lmstudio-community/embeddinggemma-300m-qat-GGUF). Nothing is sent to the cloud.

Requirements

Before you can use OmniSearch, ensure you have completed these steps:

  • Install Ollama and pull the embedding model: ollama pull hf.co/lmstudio-community/embeddinggemma-300m-qat-GGUF (see Getting Started → Setting Up Ollama).
  • In Snipset, go to Settings → AI & Models → AI Configuration and confirm Online status and that the Embed Model is set to the embedding model you pulled above.
  • Run the initial Embedding Sync (next section). Until embeddings exist, OmniSearch falls back to keyword search and shows "Ollama offline — Semantic search disabled" if Ollama is unreachable.

Running the Embedding Sync

Before Snipset can search your library by meaning, it needs to generate an embedding for every existing snippet.

  • Navigate to Settings → AI & Models → Semantic Search Sync (and/or Settings → AI Configuration) inside the main Snipset window.
  • Click Start Embedding. Snipset will process each snippet one by one. A progress bar shows how many snippets have been embedded.
  • Background Processing: The sync runs silently in the background. You can continue using Snipset normally.
  • Pause/Resume: You can click Pause to stop the process mid-way (if you need to free up computer resources) and click Resume later.
Force All: Clicking Force All deletes all existing embeddings and regenerates them from scratch. Use this only if you change your embedding model or perform massive bulk edits to your library. Otherwise, Snipset automatically updates a snippet's embedding whenever you edit it.

Using OmniSearch

Once synced, you can use OmniSearch on the dedicated OmniSearch page, or by toggling the "Semantic Mode" button inside the main Snippets search bar.

  • Type a natural language description of what you need (e.g., "apology for delay" or "my contact info").
  • Results are ranked by Relevance Score (shown as a percentage next to each result).
  • Click a result to open it in the snippet editor, or use keyboard shortcuts to insert it directly.

Here is a quick comparison of when to use which search method:

Feature Keyword Search OmniSearch
Finds exact matches ✅ Yes ✅ Yes
Finds synonyms/concepts ❌ No ✅ Yes
Works without Ollama ✅ Yes ❌ No
Speed Instant ~1–2 seconds
Requires Syncing ❌ No ✅ Yes
← PreviousClipboard HistoryNext →Statistics & Achievements