henrydaum/second-brain

What it solves

Second Brain is a personal AI assistant that runs locally on your machine, allowing it to interact with your files, shell, and local data. It aims to provide a unified interface for managing personal knowledge and automating tasks across various modalities (text, image, audio, video) using LLMs.

How it works

The system operates as a modular agentic framework. It uses an LLM loop to process requests and can be extended via packages. Key components include:

• Attachment and File Parsing: Processes various file types including PDFs, Office documents, images, audio, and video.
• Task Management: Handles both path-driven and event-driven tasks.
• Modular Capabilities: Users can install bundles like bundle_knowledgebase for RAG (indexing and searching files), bundle_memory for durable markdown-based memory, and bundle_gmail for email integration.
• Interfaces: Accessible via a terminal REPL, a Telegram bot, or a dedicated React-based Web UI.

Who it’s for

Individuals who want a powerful, locally-hosted AI assistant that can access their local machine's data and perform tasks without relying entirely on a cloud-based service.

Highlights

• Local Execution: Runs on the user's machine with support for Docker.
• Extensible Architecture: Capability bundles can be added via /packages to expand the agent's skills.
• Multi-modal Support: Capable of parsing and indexing a wide range of file formats (OCR, transcription, embeddings).
• Multimodal LLM Access: Integrates with providers like Atlas Cloud to access hundreds of curated models.
• Flexible Frontends: Supports REPL, Telegram, and a ChatGPT-style web interface.