Nova User Guide

Master Nova's AI-powered writing features for seamless document editing in Obsidian

Complete documentation for selection-based editing, chat commands, document context, and advanced features

Nova is an AI writing partner for Obsidian that enables direct, in-place editing of your documents. Unlike traditional AI tools that require copying and pasting, Nova writes exactly where you want it—right in your notes with real-time streaming updates.

Getting Started

First-Time Setup

1. Configure AI Providers

Nova supports multiple AI providers. You'll need to set up at least one:

Claude (Anthropic)
  • Claude Opus 4.6, Sonnet 4.6, Haiku 4.5
  • API key from console.anthropic.com
OpenAI
  • GPT-5.4 Pro, GPT-5.4, GPT-5.3 Chat, GPT-5.2, GPT-5.1, GPT-5 Pro, GPT-5 Mini, GPT-5 Nano
  • API key from platform.openai.com
Google AI (Gemini)
  • Gemini 3.1 Pro, 3.1 Flash-Lite, 3 Flash, 2.5 Pro, 2.5 Flash, 2.5 Flash-Lite
  • API key from Google AI Studio
Ollama (Local)
  • Any locally installed model
  • No API key needed

To add API keys:

  1. Open Obsidian Settings → Community Plugins → Nova
  2. Select your preferred provider from the dropdown
  3. Enter your API key in the corresponding field
  4. Test the connection by starting a conversation

2. Understanding Nova's Core Philosophy

Nova operates on a direct editing principle:

Selection-based editing Right-click selected text for quick transformations
Cursor-based chat AI writes exactly at your cursor position
Streaming updates See text appear in real-time as AI generates it
Writing analysis Local readability scoring, passive voice detection, and more — no AI calls needed
No copy-paste workflow Nova edits your document directly

3. Your First Edit

Try this quick example: Write a sentence in any note ("The cat sat on the mat."), select it, right-click and choose "Improve Writing," then watch Nova enhance it in real-time.

Selection-Based Editing

Selection-based editing is perfect for quick transformations of existing text. Simply select any text and right-click to access Nova's transformation menu.

Core Transformations

1 Improve Writing
Purpose: Makes text clearer, more concise, with better flow
Best for: Rough drafts, unclear sentences, verbose writing
Before "The thing that I wanted to say is that the meeting was not very productive"
After "The meeting was unproductive"
2 Make Longer
Purpose: Expands ideas with more detail and examples
Best for: Brief notes that need elaboration, outline expansion
Before "Exercise is good for health"
After "Regular exercise provides numerous health benefits, including improved cardiovascular function, stronger bones and muscles, better mental health through endorphin release, and enhanced immune system function"
3 Make Shorter
Purpose: Condenses text to essential points
Best for: Long paragraphs, wordy explanations, summary creation
4 Change Tone
Purpose: Adjusts writing style and tone
Formal Casual Academic Friendly
5 Challenge This
Purpose: Critical analysis of your writing's logic and claims — sent to sidebar chat
Best for: Checking arguments, finding logical gaps, stress-testing ideas before publishing

Unlike other transformations, this doesn't edit your text in-place. It sends a critical review to the Nova sidebar, identifying unsupported claims, logical gaps, missing counter-arguments, and questions a skeptical reader would ask.

6 Custom Instructions
Purpose: Give specific instructions for any transformation
Best for: Unique requirements not covered by other options
Examples "Rewrite this as bullet points" · "Make this sound more confident"
  "Convert to a numbered list" · "Add more technical detail"

When to Use Selection vs Chat vs Smart Fill

Selection

  • Existing text needs improvement
  • Quick, standard transformations
  • Working on mobile
  • Final polish edits

Chat

  • Adding new content
  • Combining multiple operations
  • More control over editing

Smart Fill

  • Template-based documents
  • Batch-generate sections
  • Placeholder workflows

Chat Commands & Cursor-Based Editing

Nova's chat system is designed around cursor-based editing with intelligent intention detection. Place your cursor where you want Nova to work and describe what you need—Nova automatically determines whether you want a conversational response or direct document editing.

How Intention Detection Works

Nova analyzes your message to understand your intent:

Editing Commands

Nova edits your document directly:

  • "add a conclusion here"
  • "insert a methodology section"
  • "create a methodology section"
  • "add tags: productivity, writing"
Chat Responses

Nova responds conversationally:

  • "what should I write about?"
  • "explain the difference between X and Y"
  • "help me brainstorm ideas"
  • "what's the best approach for this?"

Cursor-Based Document Editing

When Nova detects an editing intent, it writes directly at your cursor position with real-time streaming updates.

Adding New Content

Place your cursor and describe what you want:

  • add a paragraph about the benefits of exercise
  • create an introduction for this chapter
  • write a conclusion summarizing the key points
  • insert a methodology section here
  • add literature review using [[Research Notes]]
  • create dialogue showing tension between these characters
  • add error handling examples from [[API Guidelines]]
  • write a detailed analysis of the current market trends
  • develop a comprehensive project timeline
  • create a character backstory with psychological depth

Nova can generate sophisticated content at your cursor position, from simple paragraphs to complex sections that incorporate context from referenced documents. The AI draws on its training to create detailed, contextually appropriate content based on your specific request.

Document Operations

Nova can modify document metadata and properties:

  • add tags: productivity, writing
  • set status to complete
  • update the title property
  • clean up tags

Tag and Metadata Commands

Nova includes specialized commands for document properties:

Tag Management

add tags: productivity, writing, AI Adds specific tags to the document
add suggested tags Analyzes content and suggests relevant tags
clean up tags Removes duplicates and optimizes tag structure
remove tag: draft Removes specific tags from the document

Frontmatter Operations

set status to complete Updates frontmatter properties
add creation date Adds date fields to document metadata
update the title property Modifies document properties in frontmatter

What Tag Commands Edit:

  • Current document's frontmatter - Tags and properties in the active file
  • Document metadata - YAML frontmatter fields
  • Other documents - Commands only affect the current file
  • File names or folders - Only document content and properties

Protected Fields — Nova Won't Edit These

Creation dates: created, date-created, created-date, creation-date

Modification dates: modified, last-modified, updated, date-modified

Identifiers: id, uuid, uid, permalink, url, link

Protection is case-insensitive (Created, CREATED, Date-Created are all protected).

Won't create new frontmatter For general metadata operations (only exception is for tags)
Won't add properties without existing frontmatter Documents need existing frontmatter for property edits
Won't edit other documents Only the current active file can be modified
Won't modify file names or folders Only document content and properties

Natural Language Flexibility

Nova understands natural language variations. These all work for adding content:

  • "add a conclusion"
  • "write a summary at the end"
  • "create a new section about results"
  • "I need a paragraph explaining the process"

For selection-based custom instructions, use the "Custom prompt" menu option for maximum flexibility with existing text.

Undo Integration

Nova integrates seamlessly with Obsidian's native undo system:

  • First Ctrl/Cmd+Z: Removes Nova's addition
  • Second Ctrl/Cmd+Z: Restores original content (if text was replaced)
  • Works exactly like any other Obsidian edit
  • No special Nova-specific undo needed

When Chat Responds vs. Edits

Nova Edits Your Document
  • Specific instructions about content creation
  • Cursor placed where new content should go
  • Action words: add, create, write, edit, fix, improve, rewrite
Nova Responds in Chat
  • Questions or advice-seeking
  • Explanations or clarifications
  • Brainstorming or discussing ideas
  • Inquiry words: what, why, how, should, could, explain

Smart Fill

Smart Fill enables AI-powered placeholder completion for structured documents. Insert special placeholders with instructions anywhere in your documents, then generate content using the command palette, right-click context menu, the /fill command in the Nova sidebar, or the sidebar fill button.

Placeholder Syntax

Smart Fill uses HTML comment syntax for placeholders:

Basic Format:

<!-- nova: your instruction here -->

Single-Line Examples

<!-- nova: write a brief introduction -->
<!-- nova: summarize the key findings -->
<!-- nova: add a code example in Python -->
<!-- nova: explain the methodology -->

Multiline Instructions

For complex requirements, use multiline placeholders:

<!-- nova:
Write a detailed analysis that:
- Covers market trends
- Includes data from Q4 2025
- References the methodology section
-->

Insert Placeholder Command

You can also insert a placeholder using the "Nova: Insert smart fill placeholder" command, available via the command palette (Cmd/Ctrl+P) or the right-click context menu. This inserts a blank <!-- nova: --> template at your cursor so you can type your instruction directly.

Generating Content

Nova provides four ways to fill placeholders:

1. /fill Slash Command

Type /fill in the Nova sidebar. Processes placeholders sequentially top to bottom with real-time streaming.

2. Command Palette

Cmd/Ctrl+P → search "Nova: Smart fill (/fill)" → Enter. Great for keyboard-focused workflows.

3. Right-Click Menu

Right-click anywhere → "Nova: Smart fill (/fill)". Fills all placeholders. Convenient for mouse-driven workflows.

4. Margin Indicator

Click the 📝 icon next to any placeholder. Fills that specific one immediately — the most precise method.

Key Features

Contextual Awareness

Reads surrounding content, adapts tone and style, references earlier sections, and maintains document structure.

Sequential Processing

Top-to-bottom execution ensures logical flow. Earlier fills inform later ones, creating coherent connected sections.

Real-Time Streaming

Content appears progressively as it's generated. Cancel anytime with Cmd/Ctrl+. or undo with Cmd/Ctrl+Z.

Visual Feedback

📝 icons in the margin, active placeholder highlights, progress indication, and clear completion states.

Writing Effective Instructions

Do: Be Specific and Clear

<!-- nova: write a 3-paragraph introduction explaining the benefits of AI writing assistants -->

<!-- nova: create a comparison table of the top 5 features -->

<!-- nova: add Python code example showing how to implement authentication -->

Don't: Be Vague or Ambiguous

<!-- nova: add something here -->

<!-- nova: write stuff -->

<!-- nova: continue -->

Best Practices:

  1. Specify length: "Write a 2-paragraph summary" vs "Write a summary"
  2. Define scope: "Explain the authentication flow" vs "Explain this"
  3. Set tone: "Write a formal analysis" vs "Write an analysis"
  4. Include examples: "Add a Python code example using FastAPI"
  5. Reference context: "Summarize the findings from the methodology section above"

Canceling and Undo

Stopping Generation

Cancel placeholder fills at any time:

  • Keyboard shortcut: Cmd+. (Mac) or Ctrl+. (Windows/Linux)
  • Stop button: Click the stop button in Nova's interface
  • Partially generated content is preserved
  • Placeholder comment remains if generation was incomplete

Undoing Fills

Smart Fill integrates with Obsidian's undo system:

  • First Cmd/Ctrl+Z: Removes the generated content
  • Second Cmd/Ctrl+Z: Restores the original placeholder
  • Works like any other edit in Obsidian

Error Handling

If placeholder generation fails:

  • Placeholder preservation: The original placeholder comment stays in place
  • Error notification: Nova displays a clear error message
  • Continuation: When using /fill, remaining placeholders continue processing
  • Retry option: Simply run /fill again or click the 📝 icon

Common Issues:

  • API errors: Check your provider settings and API key
  • Rate limiting: Wait briefly and retry
  • Context too large: Reduce document size or remove document references
  • Network issues: Check internet connection

Best Practices

Document Structure

  • Use Smart Fill for repetitive sections across similar documents
  • Place placeholders in logical order (top to bottom)
  • Group related placeholders together
  • Use descriptive instructions that reference surrounding context

Template Workflows

Create reusable templates with Smart Fill placeholders:

  • Meeting notes: Placeholders for agenda, discussion points, action items
  • Project docs: Sections for overview, requirements, technical details
  • Research papers: Introduction, methodology, findings, conclusion
  • Blog posts: Hook, main points, examples, call-to-action

Iterative Refinement

  1. Start with broad placeholder instructions
  2. Generate initial content with /fill
  3. Review and refine the generated sections
  4. Use selection-based editing or chat commands for polish
  5. Save successful patterns as templates

Integration with Templater

Smart Fill works seamlessly with the Templater plugin:

  • Template creation: Add Smart Fill placeholders to Templater templates
  • Dynamic content: Combine Templater variables with Smart Fill instructions
  • Workflow: Insert template → Fill Templater variables → Run /fill

Example Template

---
title: <% tp.file.title %>
date: <% tp.date.now("YYYY-MM-DD") %>
---

# <% tp.file.title %>

<!-- nova: write an introduction based on the title -->

## Key Points
<!-- nova: generate 3-5 key discussion points -->

Performance Considerations

Document Size

  • Smart Fill works best with documents under 10,000 words
  • Larger documents may process slower or hit context limits
  • Consider splitting very large documents into sections

Placeholder Limits

  • No hard limit on placeholder count
  • Processing 5-10 placeholders at once is typically optimal
  • Very large batches may take longer or hit rate limits
  • Use margin indicators to fill placeholders individually for fine control

Rate Limits

  • AI providers have different rate limits
  • If rate limited, wait briefly and retry
  • Consider spreading large fill operations over time

Settings

Smart Fill can be configured in Nova settings:

  • Enable/Disable: Toggle Smart Fill feature on or off
  • Margin indicators: Show or hide 📝 icons in the editor
  • Auto-remove placeholders: Automatically remove placeholder comments after successful fill
  • Confirmation prompts: Require confirmation before filling multiple placeholders

Access settings: Obsidian Settings → Community Plugins → Nova → Smart Fill

Troubleshooting

Placeholders Not Detected

Solutions:

  • Verify exact syntax:
<!-- nova: instruction -->
  • Ensure "nova:" prefix is present with colon
  • Check for typos in the comment syntax
  • Make sure Smart Fill is enabled in settings

Margin Icons Not Appearing

Solutions:

  • Check Nova settings for margin indicator toggle
  • Restart Obsidian to refresh the editor
  • Verify placeholders use correct syntax

Generated Content Doesn't Match Instructions

Solutions:

  • Make instructions more specific and detailed
  • Add context references to surrounding sections
  • Try a different AI provider or model
  • Use Cmd/Ctrl+Z to undo and refine the placeholder instruction

/fill Command Not Working

Solutions:

  • Ensure at least one valid placeholder exists in the document
  • Check AI provider configuration and API key
  • Verify internet connection
  • Try clicking individual 📝 margin icons instead

Quick Reference

Create a placeholder:

<!-- nova: instruction -->

Or use the "Nova: Insert smart fill placeholder" command (command palette or right-click) to insert a blank template at your cursor.

Fill all placeholders (4 methods):

  • Type /fill in the Nova sidebar chat
  • Command palette: "Nova: Smart fill (/fill)"
  • Right-click context menu: "Nova: Smart fill (/fill)"
  • Click 📝 icon in margin (fills specific placeholder)

Cancel generation:

Cmd/Ctrl + .

Undo fill operation:

Cmd/Ctrl + Z

Writing Analysis

Nova includes a deterministic writing analysis engine that runs entirely locally — no AI calls, no API costs. It provides real-time feedback on your writing quality directly in the editor. For a vault-wide view, see the Writing Dashboard.

What It Analyzes

Readability grade Flesch-Kincaid grade level with a plain-language label (e.g., "Grade 7 — easy to read")
Sentence length Highlights long sentences (25+ words) and very long sentences (40+ words) with color-coded underlines
Passive voice Detects and highlights passive constructions
Adverb density Identifies adverb overuse throughout your writing
Weak intensifiers Flags words like "very," "really," "quite," "basically," "literally" that weaken writing

The Writing Stats Panel

A collapsible "Writing" panel appears in the Nova sidebar showing:

Metrics
  • Word count, sentence count, reading time
  • Readability grade and label
  • Passive voice percentage
  • Adverb density
  • Weak intensifier count
Controls
  • Toggle button to show/hide inline highlights
  • Analyze button to re-run on demand

Inline Highlights

When enabled, Nova underlines issues directly in the editor with color-coded severity:

Long sentences Very long sentences Passive voice Adverbs & intensifiers

Toggle highlights on/off from the Writing panel without losing the analysis data.

Settings

Configure in Obsidian Settings → Community Plugins → Nova → Writing Analysis:

  • Enable/disable writing analysis
  • Adjust long sentence threshold (default: 25 words)
  • Adjust very long sentence threshold (default: 40 words)
  • Toggle individual highlight types (long sentences, passive voice, adverbs, weak intensifiers)
  • Show/hide the stats panel

Opting Out Per Document

Add nova-analysis: false to any note's frontmatter to disable writing analysis for that specific document:

---
nova-analysis: false
---

Key Differentiator

Writing analysis is completely free — it runs locally using deterministic algorithms, not AI. No API key required, no token costs, works offline. Think of it as having ProWritingAid-style analysis built right into Obsidian.

Writing Dashboard

The Writing Dashboard gives you a vault-wide view of your writing quality. Open it from the command palette: Nova: Open writing dashboard.

Composite Score

Every note with at least 50 words gets a composite score from 0 to 100, built from four pillars:

Clarity 0–25

How close your readability grade is to your target level. Default target is grade 8.

Conciseness 0–25

Penalizes passive voice and very long sentences. Tighter writing scores higher.

Variety 0–25

Rewards natural sentence length variation. A mix of short and long reads better than uniform length.

Discipline 0–25

Flags adverb overuse and weak intensifiers. Show, don't tell.

Excellent 80–100 Good 60–79 Fair 40–59 Needs Work 0–39

Dashboard Layout

The dashboard has three sections:

Summary cards Vault average score (with trend arrow), total documents, analyzed count, and word count.
Trends & pillars Sparkline charts showing your composite score and passive voice percentage over time. Pillar breakdown bars show vault-wide averages for each scoring dimension.
Document table Sortable, filterable list of every note. Click a column header to sort. Click a document name to open it. Hover any metric for an explanation of how it's scored.

Trend Tracking

Each time you open the dashboard, Nova takes a snapshot of your vault's writing quality. Over time, this builds a trend line — "Your passive voice dropped 4% this month" is the kind of feedback that makes writing habits stick. Open the dashboard regularly to build your trend data.

Settings

Configure in Obsidian Settings → Community Plugins → Nova → Writing Dashboard:

  • Excluded folders — Skip folders like templates/, daily/, or archive/ from analysis.
  • Target readability grade — The grade level your writing aims for (default: 8). Academic writing may target 12–14.

Caching & Performance

The dashboard caches analysis results. On repeat visits, only files that changed get re-analyzed — a 500-file vault loads in seconds. Use the rescan button to force a full re-analysis if needed.

Opting Out

The same nova-analysis: false frontmatter flag that disables per-document writing analysis also excludes a note from the dashboard.

Document Context System

Nova can work with multiple documents simultaneously by referencing them in your conversations or by using the Auto-Context system.

Auto-Context

Nova's Auto-Context system automatically pulls in relevant notes from your vault as you write.

Outgoing Links Nova reads all notes linked in your current document
Backlinks When enabled, Nova also reads notes that link back to your current document
Toggle Enable or disable Auto-Context at any time from the Nova sidebar

Live wikilink tracking (v1.4): Adding or removing [[wikilinks]] in your document now updates the context panel automatically — no need to switch files.

Quick Panel

The Quick Panel in the Nova sidebar provides a real-time view of your current context.

Collapsed/Expanded Click the context header to toggle the visibility of active documents
Context List See exactly which notes are being used as context for the current turn
Manual Toggles Temporarily disable specific notes from the context without removing links

Document Reference Syntax

Basic Reference: [[document name]]

Manual references still work perfectly alongside Auto-Context. If you mention a note using wikilinks, Nova prioritizes that content.

Compare this section with [[Project Requirements]] Reference another note for comparison
Use the methodology from [[Research Notes]] to improve this Pull context from a linked document

Managing Context

Adding Documents Type [[document name]] anywhere in your message
Context Panel Shows which documents are currently in context
Context Limitations Each AI provider has different context limits

AI Provider Management

Supported Providers

Claude (Anthropic)
  • Models: Claude Opus 4.6, Sonnet 4.6, Haiku 4.5
  • Context: 200K tokens
  • Best for: Complex reasoning and analysis
  • Setup: API key from console.anthropic.com
OpenAI
  • Models: GPT-5.4 Pro, GPT-5.4, GPT-5.3 Chat, GPT-5.2, GPT-5.1, GPT-5 Pro, GPT-5 Mini, GPT-5 Nano
  • Context: 128K–1M tokens
  • Best for: Balanced performance and creativity
  • Setup: API key from platform.openai.com
Google AI (Gemini)
  • Models: Gemini 3.1 Pro, 3.1 Flash-Lite, 3 Flash, 2.5 Pro, 2.5 Flash, 2.5 Flash-Lite
  • Context: Up to 1M tokens
  • Best for: Fast responses and large documents
  • Setup: API key from Google AI Studio
Ollama (Local)
  • Models: Any locally installed model
  • Context: Varies by model and resources
  • Best for: Complete privacy and offline use
  • Setup: Install Ollama locally, no API key needed

Advanced Features

Streaming Text

Real-time streaming with Magical Scroll — text appears letter-by-letter as the AI writes. Your view auto-scrolls to follow the cursor for a smooth, uninterrupted experience.

Tag Management

Specialized commands: "add suggested tags" for AI-powered tagging, "clean up tags" to remove duplicates, or "add tags: productivity, writing, AI" for specifics.

Mobile Optimization

Touch-friendly selection menus, command palette for easier entry, and an optimized interface for smaller screens.

Troubleshooting

Common Issues and Solutions

Nova doesn't respond to commands
  1. Check provider settings and API key
  2. Test internet connection
  3. Try switching providers
  4. Wait if rate limited, then retry
Text appears garbled or incomplete
  1. Stop generation and retry
  2. Reduce context by removing document references
  3. Switch to provider with larger context limits
  4. Check network stability
Writing analysis not showing
  1. Check Settings → Nova → Writing Analysis → ensure "Enable writing analysis" is on
  2. Check the note's frontmatter for nova-writing: false
  3. Switch to editing/source mode
  4. Click the "Analyze" button in the Writing panel to force a refresh
Selection menu doesn't appear
  1. Ensure text is clearly selected
  2. Restart Obsidian
  3. Disable other AI plugins temporarily
  4. Check for plugin updates

Best Practices

Start with selection-based editing Quick improvements to existing text
Use chat commands for new content Place cursor and describe what you need
Reference documents strategically Add context only when needed to keep responses focused
Monitor token usage Avoid hitting provider limits by keeping context lean
Experiment with different providers Find what works best for your writing style
Use writing analysis during editing Write freely first, then let the analysis guide your polishing pass
Combine AI editing with local analysis The analysis catches the problems; "Improve Writing" fixes them

Ready to Master Nova?

Start using Nova's powerful AI writing features in your Obsidian workflow