VaultMind (Version 3.1) 🧠

VaultMind transforms Obsidian into an intelligent AI workspace. It bridges your local knowledge base (RAG) with advanced LLMs, offering a native, mobile-first experience for iOS and Desktop.

Feature Highlights:

• Smart Edit Mode: Context-aware content rewriting and appending.
• Deep Retrieval: Semantically search your vault to answer questions based on your notes.
• iOS Native: Designed to feel like a built-in Apple app.

(Note: Please replace this link with actual screenshots after deployment)

---

✨ Core Features

1. � Smart Edit Mode (New!)

VaultMind intelligently detects your intent to safe-guard your data.

• Append Mode (Default):
  • Usage: Just ask AI to "Continue writing", "Summarize the above", or "List action items".
  • Behavior: AI safely adds new text to the end of the current note. Zero risk of overwriting.
• Rewrite Mode:
  • Trigger: Use keywords like "Rewrite", "Modify", "Fix", "Optimize", or "修改".
  • Behavior: AI proposes a full replacement for the file content. You'll see a Diff preview to "Apply" or "Reject".

2. 📚 Semantic Search (RAG)

Chat with your second brain. VaultMind indexes your vault locally.

• Ask Your Notes: "What did I write about specific_topic?"
• Smart Context: Automatically injects relevant note chunks into the conversation.
• Privacy: Your index is stored locally in .obsidian/plugins/vaultmind/data.

3. 📱 Mobile-First Design

• Floating Input: A sleek input capsule that maximizes screen real estate.
• Haptic UI: Large touch targets, simplified headers, and hidden complexity on mobile.

---

⚙️ Configuration Guide

1. API Setup (Required)

Go to Settings > VaultMind.

• API Key: Enter your OpenAI / DeepSeek / OpenRouter API Key.
• Base URL:
  • OpenAI: https://api.openai.com/v1
  • DeepSeek: https://api.deepseek.com/v1
  • (Compatible with any OpenAI-format provider)
• Chat Model: e.g., gpt-4o, deepseek-chat, claude-3-opus (via proxy).

2. Embedding Model (For Search)

To enable "Ask Your Notes", you must configure the Embedding Model.

• Embedding Provider: Currently supports OpenAI-compatible embedding endpoints.
• Embedding Model Name:
  • Recommended: text-embedding-3-small (Fast, cheap, efficient).
  • Alternative: text-embedding-ada-002.
• Note: If you change the model later, you must click "Rebuild Index" to regenerate vectors.

3. Advanced Settings

• RAG Max Chunks: How many note fragments to retrieve (Default: 5). Higher = more context but higher cost/latency.
• Index Refresh: The plugin updates the index on startup. You can manually trigger it from the command palette: VaultMind: Rebuild Index.

---

� Installation

1. Download the latest release (main.js, manifest.json, styles.css).
2. Create a folder: .obsidian/plugins/vaultmind.
3. Move the files into that folder.
4. Reload Obsidian and enable VaultMind in Community Plugins.

---

🔒 Privacy & Network

• Network Request: This plugin uses Obsidian's native requestUrl API for all network traffic, ensuring compatibility with iOS sandboxes and proxy settings.
• Data Handling: Your note content is sent to the LLM provider only when you explicitly chat or build the index. No data is sent to any third-party analytics.

---

Author: LLenger License: MIT