Khoj

Khoj is an open-source, self-hosted application that functions as a personal AI assistant and knowledge management system. It allows users to interact with large language models using their own private documents and real-time internet data to create a personalized second brain. The software is deployed as a server and can be accessed via a web browser, desktop application, or mobile phone.

Users can integrate Khoj into existing workflows through dedicated extensions for Obsidian and Emacs, or via WhatsApp for mobile accessibility. It supports both local LLM execution for maximum privacy and connections to cloud-based models for higher performance. The system is designed to scale from a single-user on-device installation to a hybrid or on-premises enterprise deployment, ensuring that data remains under the control of the user or organization.

Key features

• Integration with local and online LLMs including Llama3 and GPT
• Retrieval augmented generation from PDFs, Markdown, Notion, and Word files
• Creation of custom agents with specific personas, knowledge, and tools
• Semantic search for quickly locating relevant information across documents
• Automated research tasks with personal newsletters and notifications
• Support for image generation and speech-to-text capabilities
• Multi-platform access via browser, desktop, Obsidian, and Emacs
• Ability to run entirely offline on a local computer

The architecture focuses on flexibility in model selection and data sourcing. By indexing various file formats such as org-mode, PDF, and Notion pages, Khoj provides a centralized interface for querying fragmented personal data. It is intended for researchers, developers, and knowledge workers who require a private way to manage information and automate repetitive data retrieval tasks. The software includes a specialized AI coworker called Pipali that runs directly on the user's computer to assist with daily tasks.

Khoj is a versatile AI layer that bridges the gap between general-purpose LLMs and private, user-owned data, providing a self-hostable infrastructure for personal and professional productivity.