SANE - Smart AI Note Evolution (Beta)

Keep it Simple and Stupid - An intelligent Obsidian plugin that evolves your notes automatically using AI. When you add or edit a note, SANE finds the 3 most relevant notes in your vault and enhances them with smart tags, keywords, links, and summaries.

🧠 How SANE Works

SANE follows a simple but powerful approach:

1. 📝 You add/edit a note → SANE detects the change
2. 🔍 SANE finds the 3 most relevant notes → Using AI embeddings for semantic similarity
3. ✨ SANE enhances those 3 notes → Adds tags, keywords, links, and summaries via YAML frontmatter
4. 🔄 Your knowledge evolves → Notes become more connected and discoverable over time

🎯 Why This Approach?

• KISS Principle: Simple, predictable behavior that's easy to understand
• Efficient: Only processes 3 notes per change, saving time and API costs
• Non-intrusive: Adds YAML frontmatter at the beginning, preserving your original content
• Evolving: Your vault becomes smarter as you add more notes

✨ Features

🤖 Multiple AI Providers

• OpenAI (gpt-4o-mini)
• Google AI (gemini-2.0-flash)
• Grok (grok-3-latest)
• Azure OpenAI
• Local LLMs (Ollama support)

⚙️ Smart Processing

• Configurable Triggers: Immediate, delayed (10 mins after editing), scheduled daily, or manual
• Targeted Scope: Process only specific folders to maintain control
• Simple Embeddings: Full note embeddings for better context understanding
• Cost Control: Daily budget limits and real-time cost tracking

🛡️ Security First

• Privacy Warnings: Clear notices about data usage
• Backup Reminders: Prompts to backup your vault before first use
• Folder Restrictions: Limit processing to specific folders
• Local Options: Support for local LLMs to keep data private

💰 Cost Management

• Daily Budgets: Set spending limits to prevent surprises
• Real-time Tracking: Monitor usage and costs
• Provider Comparison: See estimated costs for each AI provider
• Free Option: Use local LLMs for zero-cost processing

🚀 Quick Start

1. Install SANE

• Download from Obsidian Community Plugins (coming soon)
• Or manually install from GitHub releases

2. Configure Your AI Provider

• Choose from OpenAI, Google AI, Grok, Azure, or Local
• Add your API key in settings
• Set a daily budget (recommended: $1-2 for most users)

3. Set Your Scope (Recommended)

• Create a specific folder like "Knowledge Base" or "Smart Notes"
• Set this as your target folder in SANE settings
• Only notes in this folder will be processed

4. Initialize Your Vault

• Run "Initialize: Process all notes in target folder"
• SANE will process all existing notes, treating each as new
• This builds the initial knowledge graph

5. Start Writing!

• Add or edit notes in your target folder
• SANE automatically finds and enhances the 3 most relevant notes
• Watch your knowledge evolve!

📋 Example Output

When SANE processes a note, it adds YAML frontmatter like this:

---
sane_tags: ["machine-learning", "neural-networks", "ai"]
sane_keywords: ["deep learning", "backpropagation", "training data"]
sane_links: ["Introduction to AI", "Neural Network Basics"]
sane_summary: "Comprehensive guide to understanding neural networks and their applications in modern AI systems."
sane_version: "1.0"
created_at: "2025/5/5 19:59:11"
modified_at: "2025/6/29 12:15:12"
---

# Your Original Content
Your original note content remains completely unchanged...

💸 Cost Examples

Small vault (100 notes):

• OpenAI: ~$0.15 total
• Google AI: ~$0.05 total
• Local: Free

Medium vault (1,000 notes):

• OpenAI: ~$1.50 total
• Google AI: ~$0.50 total
• Local: Free

Large vault (10,000 notes):

• OpenAI: ~$15 total
• Google AI: ~$5 total
• Local: Free

Costs are one-time for initial processing. Daily usage is typically $0.01-0.10 depending on how much you write.

🔧 Settings Overview

AI Provider Options

• OpenAI: Most capable, moderate cost
• Google AI: Good quality, lowest cost
• Grok: X.AI's latest model, experimental
• Azure: Enterprise OpenAI with your own deployment
• Local: Ollama support, completely free and private

Processing Triggers

• Immediate: Process as soon as you edit (for instant feedback)
• Delayed: Wait 10 minutes after editing stops (recommended)
• Scheduled: Process once daily at 2 AM (for batch efficiency)
• Manual: Only process when you command it

Security & Scope

• Target Folder: Limit processing to specific folder
• Privacy Warnings: Clear disclosure about AI provider data usage
• Backup Reminders: Ensure you backup before first use

🛡️ Privacy & Security

What SANE Does

• ✅ Stores API keys locally in Obsidian
• ✅ Processes only notes in your target folder
• ✅ Adds YAML frontmatter without changing your content
• ✅ Provides local LLM option for complete privacy

What SANE Sends to AI Providers

• 📤 Your note content (for AI processing)
• 📤 Content from related notes (for context)
• ❌ Never your API keys or personal data

Recommendations

• 🔄 Backup your vault first (SANE modifies files)
• 📁 Use a specific folder to limit scope
• 🏠 Consider local LLMs for sensitive content
• 💰 Set daily budgets to control costs

🏠 Local LLM Setup

For complete privacy and zero costs:

1. Install Ollama: Download from ollama.ai
2. Download Models:

   ollama pull llama2          # For text generation
   ollama pull nomic-embed-text # For embeddings
   

3. Start Ollama: ollama serve
4. Configure SANE: Set provider to "Local" and endpoint to http://localhost:11434

🆘 Support & Development

❤️ Love SANE?

• ⭐ Star us on GitHub
• ☕ Buy us a coffee
• 🐛 Report bugs
• 💡 Suggest features

🧪 Help Test & Develop

• Test Local LLMs: Help us improve Ollama integration
• Develop: Contribute code, documentation, or translations
• Community: Share your SANE workflows and tips

❓ FAQ

Q: Will SANE change my existing notes? A: SANE only adds YAML frontmatter to the beginning of notes. Your original content is never modified.

Q: What if I don't like the AI suggestions? A: Simply delete the SANE-generated YAML frontmatter. SANE prefixes all fields with sane_ so they're easy to identify and remove.

Q: How much does it cost? A: Very little! Most users spend $0.01-0.10 per day. You can set daily budgets to control costs, and local LLMs are completely free.

Q: Is my data private? A: Your notes are sent to your chosen AI provider for processing. For complete privacy, use the local LLM option with Ollama.

Q: What if the AI returns malformed responses? A: SANE includes robust response parsing that handles JSON wrapped in markdown code blocks, malformed JSON, and provides regex-based fallback extraction. If all else fails, it gracefully returns empty results rather than crashing.

Q: How can I test if my AI provider is working? A: Use the "Test AI Connection" button in settings or run the "Test AI provider connection" command. This will show you exactly what the AI returns and help debug any issues.

Q: What happens if I have a large vault? A: Set a target folder to limit processing scope. SANE is designed to handle vaults of any size efficiently.

📄 License

MIT License - Feel free to modify and distribute!

---

SANE - Smart AI Note Evolution: Because your notes should evolve as smartly as your thinking does. 🧠✨