Skip to main content

Project banner

Sidekick

Sidekick is a JetBrains Rider plugin that brings AI-assisted development directly into the IDE, powered entirely by local language models running on your hardware via Ollama or LM Studio. No cloud API calls. No data leaving your machine. No compliance officer knocking on your door.

The Problem It Solves​

Cloud-based AI coding assistants (Copilot, Cursor, Codeium) send your code to external servers. For developers working on proprietary code, in regulated industries, or under NDAs, that's a non-starter. The alternative is typically "no AI assistance at all," which is an increasingly expensive tradeoff as these tools get more capable.

Eight specialist agents orbiting the Rider IDE core — each a different shape, each a different capability.

Sidekick provides a middle ground: AI assistance that runs entirely locally, with a plugin architecture designed specifically for Rider. It's not a generic LLM chat window pasted into an editor. It understands project structure, file context, and Rider's workspace model.

What It Does​

The core is a streaming AI chat interface backed by local models. It's context-aware--it can analyze your project structure and files to inform responses, which means the conversation is about your code, not a hallucinated version of it.

The local-first fortress: your code stays behind the wall, the cloud stays outside.

Six specialized code generators handle common tasks: documentation generation, unit test scaffolding, commit message writing, variable naming suggestions, error explanations, and JSON-to-DTO conversion. Each one is tuned for its specific task rather than relying on a generic prompt. The difference matters more than you'd expect.

The agent system is where it gets ambitious. A planning agent breaks down complex tasks and routes them to eight specialist agents: architect, implementer, reviewer, tester, documenter, debugger, optimizer, security. Reflection and critique mechanisms let agents evaluate each other's output. It's multi-agent orchestration running on your local hardware, which is either impressive or reckless depending on your GPU.

Beyond the AI features, Sidekick includes visual enhancements that don't require a model at all: depth-coded tabs, rainbow scopes for nested blocks, method separators, file age indicators, and git commit heatmaps. Quality-of-life improvements that make the IDE more readable regardless of whether you use the AI side.

Quality analysis covers exception hunting, TODO tracking, performance linting, and dead code detection. Security features include path canonicalization, shell metacharacter detection, prompt injection guardrails, and audit logging--because AI outputs are untrusted until proven otherwise.

Technology​

  • Language: Kotlin 2.0
  • Platform: JetBrains Rider 2024.1–2025.3.x
  • Build: Gradle 9.3.1, JDK 17+
  • AI Backends: Ollama, LM Studio (local only)

Streaming AI flow: code context flows into the local AI node, response tokens stream back into the IDE.

How It Connects​

Sidekick is the IDE companion for the C#/.NET development that happens across every other project. It's also an exercise in building a non-trivial AI tool with privacy as a hard constraint--a design philosophy that informs how Lexichord thinks about enterprise deployment, where data sovereignty isn't optional.

Where to Find It​