Agent-Nuvira for VS Code

This is not just a VS Code extension — it's the full Agent-Nuvira development agent, on your editor and on the CLI.

Agent-Nuvira is a complete, open-source multi-agent AI coding assistant that runs entirely from your terminal (agent-nuvira on npm — npx agent-nuvira). The VS Code extension is the editor surface for the same 17-agent engine: every agent, every routing decision, every vector-store lookup in this extension is powered by the CLI, and every feature below is equally available — and more powerful — from the command line.

# The full agent, no editor required
npx agent-nuvira execute "implement user authentication with JWT"

---

Features

🤖 17 Specialized Agents

Agent-Nuvira orchestrates 17 AI agents that collaborate to accomplish complex coding tasks — in VS Code and from the CLI:

🎯 Smart Code Assistance

• Inline Code Suggestions — As-you-type completions powered by AI agents (debounced, context-aware)
• Quick Fix — Apply agent-powered fixes to the current file
• Code Review — Full file review with actionable suggestions
• Explain Code — Get detailed explanations of selected code
• Generate Tests — Automatically create unit tests for files or selections

🔌 17+ Multiple AI Providers

Supports 17+ providers (5 built-in + 12 configurable via env vars) plus custom plugins:

• Local — Ollama, HuggingFace, GGML, LM Studio (fully offline, free)
• Groq — Fast cloud inference, generous free tier
• NVIDIA NIM — Enterprise-grade models
• Google Gemini — Free tier available
• OpenRouter — Access to 100+ models
• OpenAI · Anthropic · Mistral · Cohere · Together · DeepInfra · Fireworks · Perplexity · Azure · LM Studio · Anyscale · vLLM — configurable via environment variables
• Auto routing — a Thompson-sampling bandit learns which provider/model wins per task type and picks the best one automatically (free/local-first via the quota ledger), with mid-session failover when a key expires or a quota resets

🎛️ Model & Provider Switcher

• Status-bar model indicator — shows the active provider/model; click to switch
• Switch Model / Provider... — pick from all available providers (with availability status) or enable Auto routing; picking a provider lists its actual models (buff models) in a searchable quick-pick (native type-to-filter matches model names and ids — handy for providers with 100+ models like OpenRouter), so you can choose a specific one, or keep the provider default
• Chat Panel dropdown — a provider/model dropdown in the chat sidebar header switches the active provider (or enables Auto routing) without leaving the chat, and refreshes the status-bar indicator
• Check Model Health — runs agent-nuvira model health for the active provider and shows the report
• Auto routing toggle — when enabled, every AI flow routes each request to the best provider/model automatically: chat, execute, inline suggestions, code-lens actions, and diagnostic fixes. Note: inline completions are latency-sensitive — if suggestions feel slower with auto-routing enabled, pin a fast provider instead
• Quota Ledger View — a quota status-bar indicator (alert count when providers are parked) plus Agent-Nuvira: Show Quota Ledger opens a webview with free/local vs paid token usage, estimated savings, per-provider window tables (tokens, requests, time-to-reset, parked state), and the failover timeline (parked → failover → re-enabled). Live updates: the panel watches the CLI memory dir and auto-refreshes the moment a failover/park/window-reset is written by any process sharing it (CLI, chat, dashboard) — no manual Refresh needed, with a periodic 60s poll as a safety net for platforms where file watching is unreliable

🧠 FAISS Vector Store + Team Collaboration

• Local FAISS-backed vector store — semantic code search with a native FAISS backend (pure-JS fallback) that respects .gitignore; large contexts are chunked, embedded locally, and reduced to the top-k relevant chunks so free quotas stretch further
• Real-Time Team Collaboration — Git-synced shared config, memory, and review pipelines across your team
• buff memory backend --check — see which vector backend (faiss-native / faiss-ivf / json) is active on your machine

📊 Visual Feedback

• Agent Progress Panel — Real-time webview showing agent execution status, logs, and diffs
• Diff Viewer — Preview proposed changes with VS Code's native diff editor before accepting
• Status Bar — Quick access to agent commands and the active model indicator

Installation

From VS Code Marketplace

1. Open VS Code
2. Press Ctrl+Shift+X (or Cmd+Shift+X on macOS) to open the Extensions view
3. Search for "Agent-Nuvira"
4. Click Install

Or install from the command line:

code --install-extension dheerajsharma.agent-nuvira-vscode

From VSIX Package

Download the latest .vsix from the Releases page and install:

code --install-extension agent-nuvira-vscode.vsix

Prerequisites

The extension requires the agent-nuvira CLI to be installed:

npm install -g agent-nuvira

Note: The extension communicates with the agent-nuvira CLI via child process. Make sure agent-nuvira is available in your PATH, or configure the path in extension settings.

Quick Start

1. Configure a Provider

Set your preferred AI provider. The extension uses the same configuration as the CLI:

agent-nuvira config set defaultProvider groq
export GROQ_API_KEY=gsk_your_key_here

Or configure directly in VS Code settings (Ctrl+, → search "agent-nuvira").

2. Run a Goal

1. Press Ctrl+Shift+P (or Cmd+Shift+P on macOS) to open the command palette
2. Type "Agent-Nuvira: Execute Goal" and press Enter
3. Enter your goal (e.g., "Add authentication to the login route")
4. Watch agents collaborate in the progress panel

3. Switch Providers / Models

Click the model indicator in the status bar (e.g. chip Groq/llama-3.3-70b) or run Agent-Nuvira: Switch Model / Provider...:

1. Choose Auto routing to let the agent decide per task, or
2. Pick a specific provider (status icons show ✅ available / ⚠️ unreachable / ⏳ needs key)

If Auto routing is enabled in settings, the extension asks for confirmation before pinning a specific provider, since auto routing overrides pinned choices for chat/execute.

4. Use Keyboard Shortcuts

5. Right-Click in Explorer

Right-click any source file to Review File, Quick Fix, or Generate Tests.

Extension Settings

Commands

Requirements

• VS Code >= 1.85.0
• Node.js >= 18.0.0
• agent-nuvira CLI (npm install -g agent-nuvira)
• API key for at least one cloud provider, or Ollama for local inference

Development

# Clone the repository
git clone https://github.com/imdheerajKube/agent-nuvira.git
cd agent-nuvira/vscode-extension

# Install dependencies
npm install

# Compile TypeScript
npm run compile

# Watch mode
npm run watch

# Package into .vsix
npm run package

Marketplace Listings

• Visual Studio Marketplace
• Open VSX Registry

License

MIT