Report Abuse

Basic Information

Sidekick is a local-first macOS application that lets users chat with a local large language model that can access and cite information from files, folders and websites on the user’s machine. It is designed to run primarily offline while offering optional OpenAI-compatible API support for remote models. The app combines retrieval-augmented generation (RAG) with configurable domain-specific resource collections called experts so the assistant can surface relevant quotes, page numbers and sources. Sidekick targets researchers, students and knowledge workers who need contextual, private and verifiable answers. The project includes a native inference engine with support for local models and Apple Silicon optimizations, developer build instructions for macOS and guidance for signing required binaries. System requirements and a BYO API key option are documented.

Links

Categorization

App Details

Features
Sidekick provides a local inference engine and integrates with remote OpenAI-compatible APIs via a bring-your-own-key option. It supports RAG through configurable experts, persistent memory, a Deep Research agent that can read and synthesize 50–80 webpages, and function calling to run tools or perform actions sequentially. Reasoning model support includes DeepSeek-R1 and hybrid models such as Qwen3 with a toggleable reasoning mode. The app offers Canvas for editing and previewing text and code, automatic image generation via CoreML on supported macOS, advanced Markdown and native LaTeX rendering, automatic data visualizations for tables, syntax-highlighted code export, and a toolbox of utilities including an inline writing assistant, detector, diagrammer and slide studio. Performance features include llama.cpp backend, speculative decoding and optional offloading to speed generation on Apple Silicon.
Use Cases
Sidekick helps users conduct private, verifiable research by combining local document access, web search and RAG so answers include citations and clickable sources for easy verification. The Deep Research agent automates long-horizon, multi-step literature synthesis, saving time on reading and summarizing many pages. Memory personalization makes follow-up conversations more relevant to the user’s projects and preferences. Function calling enables practical automation such as drafting emails using local contacts and executing tools for math or string operations. Built-in tools like the inline writing assistant, detector, diagrammer and slide studio accelerate productivity tasks from editing to presentation creation. Offline-first design and Apple Silicon optimizations provide faster local generation while preserving user data privacy. The README documents installation and developer setup requirements for macOS Apple Silicon and minimum RAM.

Please fill the required fields*