Kaho

As PKM knowledge bases grow, users struggle to maintain connections between notes and rediscover relevant information. Existing AI features address this poorly, focusing on content generation rather than supporting the synthesis and linking processes that make note-taking valuable. This AI assistant plugin aims to assist the users' workflows while preserving those processes.

Made as part of my degree final year project.

features

feature	description	technology	location
ai agent	ai agent with access to the vault's knowledge graph. it can edit notes as well.	LLM, embeddings	sidebar
quote-in-chat	highlight text in your notes, then quote it directly into the agent chat without manually copying.	LLM	right-click menu
natural language search	search notes using natural language queries instead of exact keywords (e.g., "meeting notes about improving team productivity").	embeddings	sidebar
orphan note rescuer	lists notes with no outgoing links and suggests similar notes to help you make connections.	embeddings	sidebar
wikilink suggestions	detects phrases in your current note that match or relate to other notes, and suggests wikilinks to create.	embeddings, nlp	editor suggestion
similar notes	recommends semantically related notes as you work.	embeddings	sidebar
tag suggestions	suggests relevant existing tags based on your note's content.	embeddings, nlp	text editor
folder suggestions	suggests which existing folder a note belongs in.	embeddings, nlp	sidebar
note split suggestions	detects when a note covers multiple distinct topics and suggests splitting it into separate notes.	embeddings, nlp	sidebar
resurfacing forgotten notes	surfaces notes semantically similar to your current work that you haven't visited in a while.	embeddings, nlp, recency	sidebar
custom LLM providers	bring your own api key and choose from multiple LLM providers (claude, deepseek, etc).	LLM	settings
custom embedding model	embedding model that's fine-tuned on obsidian notes and this plugin's functionality	embeddings	implementation detail

privacy

• LLM features (like AI agent) sends data to your configured LLM provider. Privacy concious users can configure it to be local
• embeddings are entirely local

development

pnpm install
pnpm dev        # development mode with hot reload
pnpm build      # production build
pnpm test       # run tests

license

MIT