henrydaum/second-brain

Second Brain is an agentic framework that acts as an operating system, using local file intelligence, workflow automation, and LLMs to complete tasks and communicate over multiple modalities and messaging platforms.

What it solves

Second Brain is a local-first AI runtime that allows users to integrate a Large Language Model (LLM) into their personal workflows. It solves the problem of fragmented AI tools by providing a microkernel architecture where core logic is separated from extensible capabilities, allowing the AI to interact with local files, the web, and external messaging platforms like Telegram.

How it works

The system is built as a microkernel that handles the conversation state machine, session persistence in SQLite, and plugin management. It uses a "brain and body" metaphor: the kernel is the brain, and plugins (installed via a package store) are the body.

Key components include:

  • Conversation Runtime: A state machine that manages turns, participants, and resumable phases, ensuring that frontends (like REPL or Telegram) only handle transport and not business logic.
  • Plugin System: A modular architecture where tools, tasks, services, commands, and frontends can be live-loaded and updated without restarting the system.
  • Indexing Pipeline: A system that watches local directories, parses multiple modalities (text, image, audio, video, tabular), and creates a searchable SQLite knowledge base using lexical and semantic search.
  • Timekeeper: An event clock that enables cron-based scheduling for proactive subagents and background jobs.

Who it’s for

It is designed for developers and power users who want a programmable, local-first AI assistant that can index their personal data, automate recurring tasks, and be extended with custom plugins.

Highlights

  • Microkernel Architecture: Minimal core with a highly extensible package store for tools and frontends.
  • Local-First RAG: Comprehensive indexing of diverse file types (PDFs, code, spreadsheets, audio/video) with hybrid search capabilities.
  • Proactive Agency: Ability to schedule recurring subagents for briefings, reminders, and research via cron jobs.
  • Live-Loading Plugins: The agent can author, test, and live-load its own tools and plugins while running.
  • Multi-Frontend Support: Unified conversation logic shared across a terminal REPL and a Telegram bot.