Summary

Memex is an open-source, local-first AI journal application for iOS and Android designed for capturing life's fragments—text, photos, and voice—rather than traditional diary entries. It utilizes a sophisticated multi-agent AI system to automatically organize these fragments into timeline cards and generate insights, all while ensuring user data remains exclusively on the device. The application offers flexibility by allowing users to integrate their preferred Large Language Models (LLMs) from a wide range of providers.

How It Works

Memex employs a local-first, multi-agent AI architecture to process user-captured fragments (text, photos, voice). Specialized agents handle tasks like knowledge extraction, card generation, insight discovery, and conversational companionship. This approach ensures data privacy by keeping all records on-device and offers flexibility by allowing users to integrate their preferred LLM providers. Data is organized into interconnected Markdown files, facilitating archiving and future AI interactions. The core data flow involves input processing, knowledge extraction via the PKM agent, structured card generation, and insight discovery, all stored locally via SQLite and the filesystem.

Quick Start & Requirements

Installation involves cloning the repository, running flutter pub get, and pod install for iOS. The primary run command is flutter run --flavor globalDev. Prerequisites include the Flutter SDK (≥ 3.6.0), Xcode, and Android Studio. Users must configure an LLM API key for AI features. Official downloads are available on the App Store, Google Play, and as APKs from GitHub Releases.

Highlighted Details

• AI-Powered Organization: Multi-agent system automates tagging, entity extraction, and cross-referencing for structured timeline cards.
• Knowledge & Insights: Leverages P.A.R.A. methodology for knowledge organization, generating visual charts and narrative summaries of connections across records.
• AI Companion: Customizable AI characters offer auto-commentary on new cards and engage in persistent 1v1 conversations, remembering user context.
• Data Freedom & Privacy: All data is stored locally (filesystem + SQLite), with zero vendor lock-in and one-click export to Markdown files.
• Extensive LLM Support: Integrates with numerous LLM providers including OpenAI, Gemini, Claude, and local Ollama via OpenAI-compatible APIs.

Maintenance & Community

Contributions are welcomed, with a request to open an issue before submitting significant changes. The project includes a roadmap outlining future development priorities. No specific community channels (e.g., Discord, Slack) are listed in the provided documentation.

Licensing & Compatibility

This project is licensed under the GPL-3.0 License. As a copyleft license, GPL-3.0 requires derivative works to also be licensed under GPL-3.0, which may impose restrictions on integration with proprietary or closed-source software.

Limitations & Caveats

The project is under active development, with several features listed on the roadmap (e.g., video/file attachments, editable memory) not yet implemented. Unofficial OAuth integrations for LLM providers carry inherent risks. The GPL-3.0 license imposes copyleft requirements, potentially impacting integration with proprietary software.