felladrin/MiniSearch
Minimalist web-searching platform with an AI assistant that runs directly from your browser. Demo: https://felladrin-minisearch.hf.space
What it solves
MiniSearch provides a private, self-hosted search engine with an integrated AI assistant. It eliminates the need for third-party tracking, telemetry, and external API keys by allowing users to run AI models directly in their browser or via a self-hosted backend.
How it works
When a user enters a query, the app server uses a bundled SearXNG instance to aggregate results from various search engines. These results are then reranked locally using a cross-encoder model. If AI responses are enabled, an AI assistant reads the top results to generate a cited answer. This inference can happen entirely within the browser (via WebGPU or CPU) or through a configured remote OpenAI-compatible API.
Who it’s for
Users seeking a private search experience with AI-powered summaries and citations, as well as those who prefer self-hosting their search infrastructure to maintain control over their data.
Highlights
- Browser-based AI: Supports curated models (135M to 4B parameters) running on WebGPU or CPU, cached locally in the browser.
- Private by Design: No accounts or tracking; search history and chats are stored in the browser's IndexedDB.
- Verified Answers: AI responses include citations to the sources used and support follow-up questions with conversation memory.
- Flexible Backend: Compatible with OpenAI-compatible APIs (Ollama, vLLM, etc.) and the AI Horde.
- Complete Pipeline: Includes a bundled SearXNG metasearch engine and a local reranker for better result quality.
Related
- Project
- Project
- Project
- Project
- Project