Mane AI

What is Mane AI?

Mane AI is a private, on-device AI knowledge base for macOS that indexes documents, code, images, and audio and enables natural language chat with your data. It runs entirely on your machine, with no cloud traffic, no telemetry, and no accounts. The app is built with SwiftUI for a fast, native experience and is powered by Ollama for local large language models. Mane AI provides semantic search and multimodal capabilities, so you can query not just text but code, images and audio as well.

Key Features

• Unified Knowledge Base: Import folders and files to build your personal knowledge base, with automatic detection of code projects and smart indexing.
• RAG-Powered Chat: Ask questions about your documents and get AI responses with source citations, retrieving relevant context first.
• Semantic Search: Find files by meaning, not just exact keywords across documents, code, and image descriptions.
• Multimodal Support: Index and query images (with AI captions) and audio (with transcription).
• 100% Private: No telemetry, no cloud, no accounts. Your data stays on your Mac.
• Native Performance: Built with SwiftUI for a fast, native experience with smooth animations powered by Metal.
• Project Detection: Automatically detects software projects via manifest files such as package.json or Cargo.toml.
• Local Architecture: Client app in SwiftUI talks to a local backend, with a vector store used for semantic search.

How to Use Mane AI

Prerequisites and installation steps are provided to set up a complete local AI workflow on your Mac.

Prerequisites

• macOS 14+ Sonoma or later
• Ollama installed via brew install ollama
• Node.js 20+ via brew install node
• pnpm installed globally via npm install -g pnpm

Step 1: Install Ollama and pull the model

• Install Ollama: brew install ollama
• Start Ollama: ollama serve
• In a separate terminal, pull the AI model: ollama pull qwen2.5

Step 2: Download Mane AI

• Download the latest release from the Releases page on the repository
• Or build from source: clone the repo, install backend dependencies, open the Xcode project

Command examples:

• git clone https://github.com/ajagatobby/Mane-mac-app.git
• cd Mane-mac-app
• cd mane-ai-backend
• pnpm install
• Open in Xcode: open ../ManeAI/ManePaw.xcodeproj

Step 3: Run the app in a development environment

• Start Ollama: ollama serve
• Start backend: cd mane-ai-backend; pnpm start:dev
• Run the app: Open ManeAI/ManePaw.xcodeproj in Xcode and press Run

Importing Content

Click Import to add files or folders to your knowledge base. Content types are processed as follows:

• Code: Detected by manifest files like package.json or Cargo.toml and indexed with function and class signatures.
• Text: Text files are chunked and embedded for semantic search.
• Images: AI-generated captions describe visual content.
• Audio: Audio is transcribed to searchable text.

Supported File Types:

• Text: .txt, .md, .json, .yaml, .xml, .html, .css, .csv
• Code: .swift, .ts, .js, .py, .rs, .go, .java, .rb, .php
• Images: .png, .jpg, .jpeg, .gif, .webp, .heic
• Audio: .mp3, .wav, .m4a, .aiff, .flac, .ogg

Project Detection:

• Mane AI automatically detects projects by a set of manifest files such as package.json for Node/JS, Cargo.toml for Rust, pyproject.toml for Python, go.mod for Go, pom.xml for Java, pubspec.yaml for Flutter, Swift manifest, etc.

Chat Examples:

• You can ask questions about your indexed content and get concise AI responses with relevant citations.

Architecture

Mane AI is composed of a SwiftUI client and a local NestJS backend. The flow includes a local vector store for fast semantic search and a RAG interface for context-aware responses. All data remains on the device, ensuring privacy and security.

Use Cases

• Personal developer knowledge base: Index code, docs, and notes to quickly answer questions about your projects.
• Private research library: Keep your notes, PDFs and media locally and query them with natural language.
• Multimodal knowledge hub for creators: Index images and audio transcriptions to explore visual and audio content.
• Education and tutoring on macOS: Build a private knowledge assistant for courses, lectures and code samples.
• Private team knowledge base: Suitable for individuals or small teams who want to keep sensitive data on their own machine with local access control.

FAQ

Q: Is Mane AI free and open source? A: Mane AI is MIT licensed and open source on GitHub. You can download, modify and run it locally.

Q: What are the system requirements? A: macOS 14+ Sonoma or later, Ollama, Node.js 20+, and the ability to run Xcode projects for mac app development.

Q: Where does my data reside and how private is it? A: All data remains on your Mac, there is no cloud or telemetry; the app is designed for complete local operation.

Q: How do I get started? A: Download the latest release or clone the repo, install dependencies, configure Ollama and pull a model, then open ManePaw.xcodeproj in Xcode and run the app.