khanote 0.1.0

Universal research workflow kit connecting vibe coding tools with Obsidian

khanote

Your research was almost Khan-worthy. Now it is. You're welcome.

A universal research workflow kit bridging vibe coding tools and Obsidian.

中文文档

---

What is khanote?

khanote connects your vibe coding tools (Claude Code / Cursor / Codex / Gemini CLI / OpenCode) with Obsidian, using a pluggable research layer in between. SOPs are codified as skills — you only touch config.

Install

# pip
pip install khanote

# or Homebrew
brew tap almost-khan/tap
brew install khanote

# or pipx (isolated)
pipx install khanote

Quick Start

khanote init --vault ~/your-obsidian-vault --tool claude-code --researcher perplexity

Then inside your vibe coding tool, run:

/khanote.research.start AI agents

Features

Research Workflow

• Session-driven: each research run = a date-prefixed folder in your vault
• Pipeline: start → ingest sources → analyze → save to Obsidian
• Built-in researchers: Perplexity, arXiv, NotebookLM
• Knowledge graph: automatic linking between sessions via shared topics

Custom Researchers (Zero Code)

Add any HTTP REST API as a researcher through config.yaml — no Python required:

researchers:
  exa:
    type: http
    api_key: ${EXA_API_KEY}
    capabilities: [search]
    endpoints:
      search:
        url: "https://api.exa.ai/search"
        method: POST
        headers:
          Authorization: "Bearer {api_key}"
        body_template: '{"query": "{query}", "numResults": 10}'
        response_mapping:
          results: "$.results"
          title: "$.title"
          excerpt: "$.text"

Run /khanote.researcher.add for a guided setup — no flags, no code.

Feed Management

Set up recurring research feeds that automatically fetch and analyze content:

feeds:
  llm-papers:
    researcher: arxiv
    query: "large language models"
    keywords: [llm, transformer, agent]
    frequency: daily
    active: true

Manage feeds with guided commands:

• /khanote.feed.add — create a new feed or clone from existing
• /khanote.feed.list — see all feeds with status
• /khanote.feed.pause / resume / remove

SOP Prompt Templates

Capabilities without a direct API endpoint use SOP prompt templates — structured prompts that your vibe coding tool processes:

• analyze: PRISMA-style thematic synthesis with chain-of-thought
• ingest: metadata extraction from sources
• search: relevance ranking and filtering
• generate: structured research report generation

Multi-Tool Support

Initialize once per tool, share the same config:

Tool	Entry File	Skills Directory
Claude Code	CLAUDE.md	.claude/commands/
Cursor	.cursorrules	.cursor/rules/
Codex	AGENTS.md	.codex/
Gemini CLI	GEMINI.md	.gemini/
OpenCode	OPENCODE.md	.opencode/

Design Principles

1. Two ends fixed, middle swappable — vibe coding tool ↔ [researcher + feed] ↔ Obsidian
2. SOP as skills — workflows live in SKILL.md, not in your head
3. Zero-code for everyone — all operations via guided prompts, no CLI flags
4. Session-driven — each research run = an independent date-prefixed folder
5. Progressive disclosure — simple by default, powerful when you need it

Available Skills

Skill	Description
/khanote.research.start	Start a new research session
/khanote.research.ingest	Add sources to a session
/khanote.research.analyze	Run analysis on ingested sources
/khanote.research.save	Save results to Obsidian
/khanote.research.pipeline	Full pipeline in one command
/khanote.researcher.add	Add a custom HTTP researcher
/khanote.feed.add	Create a recurring feed
/khanote.feed.list	List all feeds
/khanote.update	Update skills to latest version

Name

Khan (from AlmostKhan) + note — Khan's notes / notes that conquer knowledge.

Status

Active development. Core research workflow and custom researcher/feed management are implemented. 222 tests passing.

License

TBD