Khoj is an AI personal assistant designed for your Digital Brain. It links up all your knowledge bases into one easily searchable and conversational interface. The purpose of Khoj is to function as an extension of your brain – a thinking tool that allows you to efficiently find and connect information across your disparate data sources.
Khoj sets itself apart by being transparent, accessible, and hackable. It is transparent in that it allows you to see how it works under the hood. Khoj is open source, so you can customize it to suit your needs. It is accessible through its simple setup and web interface. And Khoj is hackable in its modular architecture that makes it easy to add new plugins and data sources.
Features of Khoj
Khoj provides lightning-fast semantic search through your notes and documents. It uses offline AI models to enable natural language queries without needing an internet connection. As you type, Khoj surfaces relevant results instantly for a seamless search-as-you-type experience.
Searches are done completely offline. All of your personal data stays private on your own machine. Khoj indexes your content and generates vector embeddings to match your queries semantically instead of just doing fuzzy string matching. This allows you to search using related concepts and ideas, not just keywords.
Some examples of Khoj’s advanced search capabilities:
- Search for pages similar to a particular note, to discover related content
- Find notes that reference a concept, even if they don’t use the exact same keywords
- Surface diagrams and images related to a search query
In addition to search, Khoj offers a conversational chat interface. You can have natural, multi-turn conversations with Khoj to find answers, explore ideas, and expand your knowledge. Chat uses the GPT family of models from Anthropic to generate helpful, harmless, and honest responses.
To enable the chat feature, you need to add your OpenAI API key. Chat allows you to efficiently find answers and insights without having to manually dig through your notes. The conversational format helps iteratively develop new ideas and creatively riff off existing knowledge.
Here are some examples of how you can use Khoj’s chat:
- Ask for a summary of what you wrote about a specific topic
- Request examples to illustrate a concept you’ve learned
- Inquire about counter-arguments or critiques related to an idea
- Follow up on previous responses to go deeper on a subject
A core design principle of Khoj is modularity. Khoj makes it easy to build plugins to connect external data sources and front-end interfaces. This allows you to access your personal assistant directly within the tools you already use.
Khoj has first-party plugins for Emacs and Obsidian. The plugins allow you to search and chat with Khoj directly within these applications. For example, you can highlight text in Obsidian and right-click to generate a summary from your notes.
All of Khoj’s functionality is also exposed through a web interface and API. These make it simple to access your assistant from any device with a browser. The API allows developers to build custom integrations and clients for Khoj.
Self-Host and Cloud Plans
A key priority for Khoj is giving you control over your data. Khoj can be self-hosted on your own machines so that your information stays private. But we also recognize that managing a Python server is an impediment to easy adoption.
So Khoj will soon be available as a cloud-hosted service for personal use. This will allow you to access the benefits of Khoj without any server setup. Even if you use the cloud version, you will remain in control of what data Khoj can access.
To simplify self-hosting, the Khoj team is also working on a desktop app. This will handle all Python dependencies and configurations behind the scenes. Users will be able to launch their personal Khoj with one click.
For example, to install Khoj on your local machine you can run:
python -m pip install khoj-assistant
And to start the Khoj server:
This will launch the web interface to get started with setup and configuration. Find the full installation instruction on their website here: https://docs.khoj.dev/#/setup
Team behind Khoj
Khoj was founded by Debanjum Singh Solanky and Saba Imran. Debanjum and Saba are the core developers working to build Khoj into an indispensable thinking tool.
The team recognized how difficult information discovery had become with knowledge scattered across apps and devices. Khoj aims to solve this by privately connecting your data so you can efficiently find and develop ideas.
Khoj’s mission is to make your information accessible and useful regardless of where it lives. It aims to be your personal thinking assistant to help you leverage your collective knowledge.
Khoj is actively inviting users to join Discord and share their perspectives. Your feedback is crucial for improving Khoj and driving development priorities. Together, we can evolve Khoj into an AI assistant that feels like a natural extension of your mind.
Some key ways you can help contribute to Khoj:
- Give product feedback and feature requests based on your needs
- Try out new features early and report any bugs
- Spread the word to help the community grow
- Contribute code to the open-source project on GitHub