I made Gemma4 E4B run directly inside Obsidian, it allows me to chat with any specific note or the whole vault of notes! Even better, it uses Andrej Karpathy's llm-wiki format for each note so it helps me connect ideas and organize them into a knowledge base . Everything it writes is plain Markdown, but your original notes stay untouched!
Currently my creation supports:
- Turning the notes you've already collected into a living wiki.
- Summary cards + an index: See what’s in your vault at a glance.
- Find orphan notes & reconnect ideas: Discover isolated notes and hidden relationships.
- Related links + shared tags: Relink your knowledge and surface connections.
- Concept pages: Bring related notes together around emerging topics.
- Find gaps and contradictions: Surface missing pieces and conflicting claims.
- Quizzes + flashcards: Turn saved notes into something you can revisit.
LiteRTLM + WebGPU do the work, and the model runs inside Obsidian’s renderer process. So after a one-time ~3 GB model download, it all runs entirely locally — and offline with #WebAI.
If you’re a fan of Obsidian Web Clipper and have ended up collecting too much — work notes, learning notes, ideas, articles, papers, and saved posts... this is for turning that growing pile into a connected, searchable, living knowledge base.
Free, private, and local. No API key. No account. No Ollama. No LM Studio. No separate inference server. No subscription.
Two days since launch — nearly 200 downloads. Download link: https://lnkd.in/gp8s5Zp9
Let me know if there’s any feature requests or if you find any bugs. I’d really appreciate any feedback!
I made Gemma4 E4B run directly inside Obsidian, it allows me to chat with any specific note or the whole vault of notes! Even better, it uses Andrej Karpathy's llm-wiki format for each note so it helps me connect ideas and organize them into a knowledge base . Everything it writes is plain Markdown, but your original notes stay untouched!
Currently my creation supports: - Turning the notes you've already collected into a living wiki. - Summary cards + an index: See what’s in your vault at a glance. - Find orphan notes & reconnect ideas: Discover isolated notes and hidden relationships. - Related links + shared tags: Relink your knowledge and surface connections. - Concept pages: Bring related notes together around emerging topics. - Find gaps and contradictions: Surface missing pieces and conflicting claims. - Quizzes + flashcards: Turn saved notes into something you can revisit.
LiteRTLM + WebGPU do the work, and the model runs inside Obsidian’s renderer process. So after a one-time ~3 GB model download, it all runs entirely locally — and offline with #WebAI.
If you’re a fan of Obsidian Web Clipper and have ended up collecting too much — work notes, learning notes, ideas, articles, papers, and saved posts... this is for turning that growing pile into a connected, searchable, living knowledge base.
Free, private, and local. No API key. No account. No Ollama. No LM Studio. No separate inference server. No subscription. Two days since launch — nearly 200 downloads. Download link: https://lnkd.in/gp8s5Zp9
Let me know if there’s any feature requests or if you find any bugs. I’d really appreciate any feedback!