Skip to main content

Frequently Asked Questions

💡TL;DR

Quick answers to the most common Notemd questions. Covers installation (5 min setup), 30+ LLM providers support, local/cloud model options, one-click workflows, multilingual UI (21 languages), wiki-linking automation, research integration, PDF reading workflows, diagram generation, and open-source licensing. All features work offline with Ollama.

Installation & Setup​

How do I install Notemd in Obsidian?​

Open Obsidian Settings → Community Plugins → Browse → search "Notemd" → Install → Enable.

Requirements:

  • Obsidian 0.15.0+
  • Supported on Desktop (Windows, macOS, Linux) and Mobile (iOS, Android)

Manual installation:

  1. Download main.js, styles.css, manifest.json from GitHub Releases
  2. Place them in <your-vault>/.obsidian/plugins/notemd/
  3. Reload Obsidian and enable the plugin

Stats: 4k+ downloads, 44 releases, MIT licensed.

LLM Configuration​

What LLM providers does Notemd support?​

Notemd supports 30+ providers:

Cloud providers:

  • OpenAI (GPT-4, GPT-4o, GPT-4o-mini, o1, o3)
  • Anthropic (Claude Opus, Sonnet, Haiku)
  • Google (Gemini Pro, Flash, Ultra)
  • xAI (Grok)
  • Mistral, DeepSeek, Qwen, Moonshot, GLM

China-focused:

  • Qwen, Qwen Code, Doubao, Moonshot, Xiaomi MiMo, GLM, Z AI, MiniMax
  • Huawei Cloud MaaS, Baidu Qianfan, SiliconFlow

Local/self-hosted:

  • Ollama
  • LM Studio
  • Any OpenAI-compatible endpoint

Gateways:

  • Azure OpenAI, OpenRouter, Together AI, Groq, AIHubMix
  • GitHub Models, Fireworks, LiteLLM, Nebius, Cerebras
  • Hugging Face, Vercel AI Gateway, Requesty

Can Notemd work with local LLMs?​

Yes, fully supported. Notemd works with:

  • Ollama (native integration)
  • LM Studio (via OpenAI-compatible endpoint)
  • vLLM, LocalAI, text-generation-webui (via OpenAI-compatible mode)

All tasks work with local models: linking, concept generation, translation, research, and diagram generation.

Mix & match example:

  • Ollama (local) for linking → fast, private, no cost
  • OpenAI for research → high quality, web search integration
  • Gemini for translation → multimodal support

How do I configure different models for different tasks?​

In Notemd Settings, each task has independent provider/model selectors:

TaskExample Configuration
LinkingDeepSeek (fast, cheap)
Concept GenerationClaude Opus (high quality)
ResearchPerplexity or GPT-4o (web-grounded)
TranslationGemini Flash (multimodal)
Diagram GenerationOllama (local, private)

This per-task optimization balances speed, cost, and quality.

Core Features​

How does Notemd differ from ChatGPT for paper reading?​

DimensionChatGPTNotemd
Output locationChat history (ephemeral)Obsidian vault (persistent files)
Result formatText answerWiki-links, concept notes, diagrams, translations
Long-term valueDisappears after sessionAccumulates into knowledge base
ReusabilityMust re-queryFiles stay, link, evolve

Use ChatGPT for: Quick Q&A, one-off explanations Use Notemd for: Building a persistent, structured knowledge base over weeks/months

Process:

  1. You run "Process file (add links)" on a note
  2. Notemd sends the text to your configured LLM
  3. The LLM identifies key concepts in context
  4. Notemd inserts [[wiki-links]] at each concept occurrence
  5. Optionally, it auto-creates concept notes in a folder you specify

Features:

  • Synonym suppression (avoids "transformer" vs "transformers" duplicates)
  • Backlinks from concept notes to source
  • Link integrity on rename/delete
  • Pure extraction mode (extract concepts without modifying the original)

What is the Notemd one-click workflow?​

One-click workflows chain multiple actions into a single sidebar button.

Default workflow: One-Click Extract

process-current-add-links > extract-concepts-current > research-and-summarize > summarize-as-mermaid

Result: One click runs:

  1. Add wiki-links to current note
  2. Generate concept notes for new terms
  3. Research & append background
  4. Compress into a Mermaid diagram

Custom workflows: Define your own in Settings with the DSL format. Each workflow becomes a reusable sidebar button.

How does Notemd compare to Copilot or Smart Connections?​

These plugins solve different problems:

  • Copilot — Conversation-first. You chat with an AI about your notes. AI answers stay in the chat panel; they don't modify your vault.
  • Smart Connections — Discovery-first. It finds semantically similar notes using embeddings. It doesn't create or modify content.
  • Notemd — Write-first. AI processes your notes and writes structured results (wiki-links, concept notes, research, diagrams, translations) directly into your vault as persistent files.
What you needRecommended plugin
Chat with AI about notesCopilot
Find similar/related notesSmart Connections
Auto-insert wiki-links and build a knowledge graphNotemd
Generate concept notes with backlinksNotemd
Create diagrams from note contentNotemd
Research a topic and append summariesNotemd

All three can be used together — they complement rather than compete.

Research & Translation​

Which search services work with Notemd research?​

ServiceAPI Key RequiredQualityNotes
TavilyYesHighRecommended, AI-optimized search
DuckDuckGoNoExperimentalNo setup, lower quality

How "Research & summarize" works:

  1. You select a topic or run it on current note
  2. Notemd queries the search service
  3. Results → LLM summarization
  4. Summary appended to your note with source citations

Result: Persistent research trail in your vault, not browser tabs.

Does Notemd support Chinese or other languages?​

UI languages: 21+ supported

  • English, 中文 (简体), 中文 (繁體), 日本語, 한국어
  • Arabic, German, French, Spanish, Russian, Portuguese, Italian
  • Hindi, Bengali, Thai, Vietnamese, Turkish, Polish, Czech, and more

Task output language: Independent configuration

  • UI in Chinese + output in English: ✅
  • UI in English + output in Chinese: ✅

Use case: Bilingual researchers can keep UI in native language while generating English notes for publication.

PDF & Academic Workflows​

Can I use Notemd for paper/PDF reading?​

Yes. Workflow:

  1. Convert PDF → Markdown using MinerU (recommended for academic papers)
  2. Open the Markdown file in Obsidian
  3. Run Notemd tasks:
    • Add wiki-links to key terms
    • Generate concept notes for unfamiliar terminology
    • Extract original text via custom questions (e.g., "What are the limitations?")
    • Research background for cited methods/datasets
    • Translate difficult sections
    • Compress to Mermaid diagram for structural understanding

Version 1.9.1+: Chapter structure extraction with TOC support

Result: Your paper reading becomes a structured, linkable knowledge artifact in your vault.

Diagrams & Visualization​

What graph/diagram types can Notemd generate?​

TypeSupportFeatures
MermaidFullFlowchart, sequence, class, state, ER, Gantt, syntax auto-fix, batch validation
JSON CanvasFullObsidian native canvas layouts
Vega-LiteFullData charts, time series, bar, scatter, customizable
HTMLFallbackFor unsupported diagram types
Editable HTML/SVGExplicit targetSelf-contained inline SVG with semantic Draw.io-style annotations
Draw.io / Drawnix artifactsExport boundaryDeterministic XML/JSON from SemanticFigureModel; no editor runtime dependency

Mermaid auto-fix: Deep debug mode repairs:

  • Broken arrows (-->, -.->, ==>)
  • Invalid labels, comments
  • Subgraph syntax errors
  • Shape notation ([, (, {, (()
  • Pipe characters in labels
  • Note attachments

Workflow:

  1. Run "Summarize as Mermaid diagram" on a note
  2. Preview in modal
  3. Export as SVG or PNG
  4. Or save as .mmd file in vault

Open Source & Community​

Is Notemd free and open source?​

Yes.

You own:

  • Your data (stays in your vault)
  • Your workflows (customize prompts, models, chains)
  • Your privacy (local LLM support)

Contribute: Pull requests welcome. See CONTRIBUTING.md

Troubleshooting​

My LLM calls are failing. What should I check?​

  1. API key valid? Check Settings → Provider Configuration
  2. Model name correct? Use "获取模型列表" (Get Model List) button in settings
  3. Network accessible? Some providers require VPN
  4. Rate limits? Check provider dashboard
  5. Diagnostics: Run "Connection Test" in settings for detailed error logs

Notemd is slow. How can I speed it up?​

  1. Use faster models: DeepSeek, Gemini Flash, GPT-4o-mini
  2. Reduce chunk size: Settings → Smart Chunking → lower word count
  3. Parallel processing: Settings → Batch Processing → increase concurrency
  4. Local models: Ollama for tasks that don't need web search

Can Notemd work 100% offline?​

Yes, with Ollama. No API key, no internet, no data leaves your machine:

  1. Install Ollama and run ollama serve
  2. In Notemd settings, select Ollama as your provider
  3. Pull a model: ollama pull llama3 (or qwen2.5, gemma2, etc.)
  4. All features work: wiki-links, concept notes, translation, diagrams

DuckDuckGo research also works without an API key (experimental quality).

With cloud providers (OpenAI, Anthropic, Google, DeepSeek, etc.), your note content is sent as API input. Check each provider's privacy policy for data handling details.

Is my data sent to the cloud?​

It depends on your provider:

Provider typeData sent to cloud?Notes
Ollama (local)NoEverything stays on your machine
LMStudio (local)NoSame as Ollama
DeepSeek, OpenAI, etc.YesNote content sent as API input for processing
OpenRouter, AIHubMixYesGateway that routes to cloud providers

Notemd itself collects no telemetry, no analytics, no user data. The plugin is fully open source (MIT). The only data that leaves your machine is what you explicitly send to your configured LLM provider.

Which Obsidian AI plugin is best for research?​

Depends on your research workflow:

NeedPluginWhy
Find related papers/notesSmart ConnectionsSemantic similarity discovery
Chat about your sourcesCopilotConversation with vault context
Auto-link concepts + generate notesNotemdAI identifies concepts, inserts links, creates notes
Web research + summarize into notesNotemdTavily/DuckDuckGo search → LLM summary → appended to note
Generate diagrams from researchNotemdMermaid/Canvas/Vega-Lite from note content

Recommended combo: Smart Connections (discovery) + Notemd (structuring) + Copilot (discussion).

Getting Help​