Astrelium

Files
Extension demo

Inspiration

As developers, we’ve always wanted an AI assistant that truly understands our code not just snippets, but the entire project context. Most tools rely on cloud based models, which compromise privacy and often feel disconnected from the actual development flow. We built Astrelium to solve that: a local first AI coding assistant that runs entirely inside VS Code, powered by open source LLMs like gpt-oss:20b via Ollama. It’s fast, private, and feels like a real teammate one that respects your workflow and your data.

What it does

Astrelium is a fully offline AI coding assistant for VS Code that empowers developers to generate, refactor, and review code using natural language; analyze project structure and suggest architectural improvements; create documentation, tests, and migration plans; audit code for security and performance; and upload and analyze any file in the workspace all without relying on the cloud, ensuring complete privacy and seamless integration into local workflows.

How we built it

It's built with a sleek frontend using the VS Code Extension API and WebView, backed by a Node.js engine that handles file operations, context parsing, and command execution. It runs local LLMs via Ollama, finely tuned for developer workflows, and features a glass morphism UI with syntax highlighting and persistent chat history. Its natural language command system enables intuitive triggers like “Refactor this using MVC,” “Generate unit tests,” or “Audit this file for security issues,” making offline coding assistance feel seamless and powerful.

Challenges we ran into

Building Astrelium meant tackling key technical challenges head on: integrating local LLMs with real-time responsiveness, ensuring deep context awareness across sprawling codebases, and designing a UI that feels native and intuitive within VS Code. The team also had to balance performance with fully offline processing, while carefully handling edge cases in file parsing and natural language command interpretation all to deliver a seamless, private, and developer first experience.

Accomplishments that we're proud of

Astrelium is a fully offline AI assistant with zero telemetry, seamlessly integrated into VS Code and powered by Ollama for local LLM execution. It offers advanced capabilities like migration planning, API documentation generation, and automated backup creation all wrapped in a polished, modern UI designed to elevate developer productivity while preserving privacy and control.

What we learned

Building Astrelium taught us how to craft scalable VS Code extensions by modularizing features and optimizing activation flows, while integrating local LLMs required careful tuning for responsiveness and context depth. We learned that great UX in developer tools hinges on clarity, speed, and familiarity especially when designing natural language interfaces that translate technical intent into precise actions. And throughout, we prioritized secure, private AI workflows by keeping everything offline, minimizing telemetry, and ensuring full control over data and execution.

What's next for Astrelium

Astrelium’s roadmap is packed with upgrades aimed at deepening its impact: enhanced context awareness will allow smarter suggestions across large, complex codebases; expanded natural language commands like “Convert this to TypeScript” or “Generate API docs for this module” will make interactions even more intuitive; interactive model comparison tools are in development to help users benchmark and select the best local LLM for their needs; UI refinements will build on its glass morphism design for smoother, more customizable experiences inside VS Code; and its offline first security and performance audits are set to become even more robust, scanning for vulnerabilities and inefficiencies with precision.