🚀 The Motivation
Cloud AI coding assistants are convenient, but subscription fees add up and sending enterprise or personal codebases to external APIs isn't ideal for privacy.
I wanted a simple, zero-cost, local-first alternative that integrates directly into VS Code and hooks straight into Ollama. So, I built Echo.
🛠️ How It Works Under the Hood
To keep the MVP fast and reliable, the architecture is entirely decoupled from heavy frontend frameworks:
-
Extension Host (TypeScript): Listens for commands and sends direct HTTP requests to the local Ollama instance running at
http://localhost:11434. -
Inline Completion Provider (TypeScript): Hooks directly into VS Code's native inline completion API (
vscode.InlineCompletionItemProvider) to serve ghost-text suggestions as you type. -
Distribution: Packaged directly as a
.vsixbinary via GitHub Releases for complete control.
⚡ Current Challenges & What's Next
- Hardware Variability: Users running different local models need smooth onboarding. My next step is adding automatic model detection via Ollama's API on startup.
- Streamlined Workflow: Refining how code snippets and prompts are handled directly inside the editor workflow.
📦 Try It Out
Echo is available as an open-source MVP release on GitHub:
- GitHub Release: Download Echo MVP
- Requirements: VS Code + Ollama running locally.
I'd love feedback from fellow developers building local toolchains!
Top comments (0)