RAG Frameworks · johnbean393

Sidekick

Sidekick is a native macOS application that lets you chat with local LLMs while giving them access to your files, folders, and websites—all without installing additional software. It runs entirely offline using llama.cpp for inference, supports RAG-based retrieval, and optionally integrates with remote APIs like OpenAI and Anthropic.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/johnbean393/Sidekick

3.3k

GitHub stars

144

Forks

Swift

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	johnbean393/Sidekick
Owner	johnbean393
Primary language	Swift
License	MIT — OSI-approved
Stars	3.3k
Forks	144
Open issues	35
Latest release	Unknown
Last updated	2026-05-24
Source	https://github.com/johnbean393/Sidekick

What Sidekick is

Written in Swift/SwiftUI, Sidekick embeds llama.cpp for local GGUF model inference, implements RAG for document/web retrieval, supports function calling for agentic workflows, and offers optional OpenAI-compatible API integration. It includes markdown rendering with LaTeX, data visualization, and speculative decoding optimization for Apple Silicon.

Quickstart

Get the Sidekick source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/johnbean393/Sidekick.git2cd Sidekick3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Research and Deep Dives

Students and researchers can activate topic-specific 'experts' (curated file/folder collections) and ask questions with inline source citation. Deep Research agent handles multi-step synthesis across 50–80 webpages.

Local-First RAG Applications

Organizations wanting private, offline document Q&A without external APIs can deploy Sidekick as a macOS client. Data never leaves the device; local inference scales to large corpora via RAG retrieval.

Agentic Workflows on macOS

Function calling enables agents to draft emails (pulling contact data), generate financial reports, edit code in Canvas, or trigger system actions—all running locally without remote API calls.

Implementation considerations

Requires macOS development environment (Xcode, Swift toolchain) to build from source; pre-built binaries status unknown—verify availability before deployment.
Local model selection and quantization (GGUF format) must be tested for inference latency and memory footprint on target Mac hardware (especially older models).
RAG retrieval quality depends on document preprocessing, embeddings model choice, and chunking strategy—not all details exposed in README.
Function calling capability requires careful permission and sandboxing review (e.g., email draft access, file system interactions) before production use.
Web search and image generation (Apple Intelligence on macOS 15.2+) are optional; verify feature availability on your deployment macOS versions.

When to avoid it — and what to weigh

Windows or Linux Deployment Required — Sidekick is macOS-native only. No cross-platform support is evident. Team members on Windows/Linux cannot use it.
Enterprise Multi-User or Mobile Needs — No indication of server deployment, multi-user sync, or mobile clients. Suited for single-user local workflows, not enterprise dashboards or mobile-first apps.
High-Volume Real-Time API Services — Sidekick is a client application, not an API service. If you need to expose LLM+RAG endpoints to many concurrent users or integrate with backend systems, this is not the right tool.
Models Requiring Active Development/Updates — No release history provided; last push is May 2026 (recent), but no release cycles mentioned. If you depend on frequent model or feature updates, maintenance cadence is unclear.

License & commercial use

MIT License (permissive, OSI-approved). Allows commercial and private use with attribution.

MIT is a permissive license that explicitly allows commercial use, modification, and redistribution. However, you must retain the license and copyright notice. No warranty is provided. If you modify Sidekick for a commercial product, ensure you comply with the MIT terms and any upstream dependencies (llama.cpp, etc.)—review their licenses separately.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	High

Security considerations

Data stays local by default (offline inference), reducing exposure. However: (1) Function calling grants filesystem, contacts, and email access—requires explicit user consent and OS sandbox validation; (2) Web search and image generation involve network calls and Apple Intelligence APIs—review data handling; (3) No mention of encryption at rest, secure credential storage for API keys, or security audit. Assess your threat model and data sensitivity before deployment.

Alternatives to consider

LM Studio

Cross-platform (Windows, macOS, Linux) local LLM client with similar GGUF model support and UI. Broader OS coverage if Windows/Linux users are in scope.

Ollama

Lightweight command-line LLM runtime (macOS, Linux, Windows) with REST API. Better for headless/server deployments and multi-tool integration, though lacks native RAG and UI.

ChatGPT / Claude Web Apps + Self-Hosted RAG (e.g., LlamaIndex, Langchain)

Decoupled approach: use managed API services for inference and build custom RAG with open-source frameworks. More flexible for enterprise workflows but requires more engineering.

Software development agency

Build on Sidekick with DEV.co software developers

Contact Devco to evaluate Sidekick for your team, customize it for enterprise needs, or architect alternative local-first AI solutions that fit your security and platform requirements.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

sidekick llama_index llm-for-zotero

Related on DEV.co

Explore the category and the services that help you build with it.

RAG Frameworks AI Application Development Custom Software Development Talk to DEV.co

Sidekick FAQ

Can I run Sidekick on Windows or Linux?

No. Sidekick is macOS-native (Swift/SwiftUI) and not ported to other operating systems. Cross-platform alternatives like LM Studio may be suitable.

Do conversations and files stay private?

Yes, by design. Sidekick runs locally with no required external services. If you use OpenAI-compatible API integrations, data sent to those APIs is subject to their privacy policies.

What LLMs does Sidekick support?

Built-in support for GGUF-format models (Qwen, Deepseek-R1, Llama, etc.) via llama.cpp. Optional remote API support for OpenAI, Anthropic, DeepSeek, Groq, Mistral, and others via bring-your-own-key.

Is there a server/cloud version or multi-user deployment?

No indication in the README. Sidekick appears to be a single-user macOS client application only. Multi-user or cloud deployment would require custom engineering.

Software developers & web developers for hire

DEV.co is a software development agency delivering custom software development services to companies building on open source. Our software developers and web developers design, integrate, and ship production systems — spanning web development, APIs, AI, data, and cloud. If Sidekick is part of your rag frameworks roadmap, our team can implement, customize, migrate, and maintain it.

Ready to Deploy Sidekick or Build Custom AI for macOS?

Contact Devco to evaluate Sidekick for your team, customize it for enterprise needs, or architect alternative local-first AI solutions that fit your security and platform requirements.

Talk to DEV.co Schedule a Consultation