Lyra OpenCoder AI

Lyra OpenCoder AI is an autonomous AI coding assistant for Visual Studio Code — a Claude Code / Cursor-style agent that lives in your sidebar, can read and edit your project, run terminal commands (with your approval), search your codebase semantically, and handle the day-to-day busywork of coding.

It works fully offline with Ollama, or against a cloud provider (OpenAI-compatible endpoints, OpenRouter) if you'd rather use a hosted model — you choose per the Settings tab.

Features

Chat

• Streaming chat with markdown, tables, and collapsible "thinking" blocks for reasoning models
• Autonomous agent mode: the model can read files, write/edit/rename/delete files, search the codebase, and run terminal commands to complete a task end-to-end
• Insert-at-cursor / replace-selection / copy buttons on every code block
• Visible diffs on file edits and captured output on commands, right in the chat

Sidebar

• Chat — the main conversation view
• History — every past conversation, saved locally, searchable and resumable
• Files — a lazy-loading file tree; pin any file to include it in your next message's context
• Prompts — a library of built-in actions (audits, generators, git tools) plus your own saved custom prompts
• Settings — provider/model/generation settings and your memory profile

Inline actions (right-click a selection, or use the Command Palette)

• Explain, Fix, Refactor, Optimize, Convert Language
• Generate Tests, Generate Docs, Generate API Documentation
• Find Bugs, Security Audit, Performance Audit, Accessibility Audit
• Generate SQL, Generate Regex
• Quick Fix integration: errors and warnings get an "Ask Lyra to fix" code action

Code completion: inline ghost-text suggestions as you type (Tab to accept, Esc to dismiss), backed by whichever provider you've configured.

Codebase search: Lyra OpenCoder: Index Workspace builds a local SQLite index of your project, chunked by function/class/method (not just whole files), so @codebase <question> in chat retrieves relevant, cited snippets. The index stays fresh automatically as you save files.

Git integration: AI-generated commit messages (populates the Source Control input box directly), Explain Git Diff, and Review Pull Request (diffs your branch against main/master).

Memory: an editable profile (coding style, preferred language/framework, favorite libraries) that Lyra keeps in mind across every conversation — you write it, nothing is silently inferred.

Terminal: Lyra can run commands to help complete a task, but every command requires your explicit approval first — it is never executed automatically.

Prerequisites

• For fully local/offline use: Ollama installed and running. Lyra defaults to qwen2.5-coder:1.5b, but any model you've pulled works.
• For cloud use: an API key for an OpenAI-compatible endpoint or OpenRouter, set in the Settings tab.

Getting Started

1. Open the Lyra OpenCoder view from your Activity Bar.
2. Pick a provider in Settings (Ollama is the default — just make sure it's running).
3. Start chatting, highlight code and right-click for quick actions, or run Lyra OpenCoder: Index Workspace to enable @codebase search.

Privacy

Lyra never adds its own telemetry or third-party analytics. What happens to your code depends entirely on the provider you choose in Settings: with Ollama, everything stays on your machine; with an OpenAI-compatible endpoint or OpenRouter, your code is sent to that provider the same way it would be for any API call you made directly.

License

This project is licensed under the MIT License - see the LICENSE file for details.