Opportunity report

Searching by meaning — PKM needs semantic search / local embeddings

Users with thousands of notes find keyword/tag search insufficient. They either buy/try Devonthink Pro or add plugins (Smart Connections) or experiment with local LLM embeddings. The clear desire is for reliable semantic search that runs locally or privately and surfaces relevant content.

Worth it score 80
High confidence
4 signals
8 evidence

Persona

1 evidence
Users with large note collections (writers, researchers, creative users) who need recall beyond tag or keyword matching.

I might write something short... but if that sentence has a great poetic rhythm ... an app smart enough to analyze my notes and say 'hey this one matches your work' would be useful.

view evidence on reddit

Pain

3 evidence
Traditional keyword/tag search misses useful matches; users want apps that analyze content semantically and surface notes that 'match your work' (e.g., find poetic phrasing, similar ideas) across thousands of notes.

But if the note taking app is smart enough to analyze my notes and say “hey this one matches your work in this field.” Then this whole pkms become useful.

view evidence on reddit

Devonthink... Built-in AI (in the Pro version) helps searching and finds connections between documents in your filesystem.

view evidence on reddit
+1 more evidence

Smart Connections uses local embeddings and your Smart Environment to surface notes that are semantically related.

view evidence on reddit

Workaround

3 evidence
Users try DEVONthink (Pro with AI), employ Smart Connections or similar plugins, or build local LLM/embedding pipelines to surface semantically similar notes.

You could try conversational search with an LLM and Logseq with this: conversational-search-on-logseq-with-local-llms

view evidence on reddit

Smart Connections plugin (uses local embeddings) surfaces semantic matches.

view evidence on reddit
+1 more evidence

I'm working on a notes app that uses local llms to generate embeddings for each block, and you can then find blocks that are semantically similar.

view evidence on reddit

Desire

1 evidence
Semantic/meaning search that suggests relevant notes (local embeddings/LLMs or built-in AI) to surface latent connections without rigid tags.

If the note taking app is smart enough to analyze my notes and say “hey this one matches your work in this field.” Then this whole pkms become useful.

view evidence on reddit
Built by Signal Hunt · Every claim includes traceable evidence.