The words you remember aren’t always the words you copied. Meaning Search finds related text in your history using an optional model running on your device.
Install and start Ollama. In ClipsX, open Intelligence → Models and connect your local endpoint.
02
Choose an embedding model
Select an installed model that reports embedding support, then enable Meaning Search. A text-generation model serves a different purpose.
03
Let the index build
Indexing runs in the background. Exact text search remains available while your semantic index is prepared.
Common questions
A few useful details.
What if Ollama is unavailable?+
Exact text search remains available. Check that Ollama is running, refresh the connection in Intelligence, and confirm your selected model is installed.
Why doesn’t a result match what I meant?+
Semantic matches are approximate and depend on the model and indexed content. Try different wording or fewer filters. Use exact search for identifiers, commands, paths, and error messages.
What does the similarity percentage mean?+
It measures similarity in the selected model’s embedding space, not the probability that a result is correct. An optional minimum filters semantic matches only; it never removes exact text matches.
Can I change the model or delete the index?+
Yes. Changing the embedding model builds a replacement index alongside the active one. Deleting the Meaning Search index leaves your original clips and exact search intact. Ordinary clip edits update that clip’s index data.
Need an answer across several clips?
Recall uses a separately configured generation model to answer questions with inspectable citations.