Vivus

An AI coding assistant VSCode extension powered by Local AI, designed to work like Amp but with your own local/hosted language model.

Features

🤖 AI-Powered Coding Assistant: Get help with coding tasks, debugging, and explanations
🔧 Tool Calling Support: Execute file operations, terminal commands, and searches
💬 Interactive Chat Interface: Clean, integrated sidebar panel within VSCode
🌐 Local AI Integration: Connect to your Local AI instance with qwen3-4b-thinking model
📁 File System Tools: Read, write, edit files and directories
⚡ Terminal Integration: Execute shell commands directly from chat
🔍 Advanced Search: Find files and search content across your codebase
📚 Chat History: Store and retrieve conversation history with Appwrite backend
🔄 Streaming Responses: Real-time response display as Vivus generates text
⚙️ Easy Configuration: Settings button for quick access to preferences
🔁 Auto-Retry: Automatic retry logic for network timeouts and errors

Installation

Prerequisites

Local AI: Set up Local AI at https://llm.changeme.ai with the qwen3-4b-thinking model
Node.js: Version 16 or higher
VSCode: Version 1.74.0 or higher

Building from Source

Clone this repository:

git clone <your-repo-url>
cd vivus-agent

Install dependencies:
```
npm install
```
Build the extension:
```
npm run compile    # Development build
npm run package    # Production build (optimized)
```
The extension uses webpack for bundling, producing a single optimized file.
Open in VSCode and press F5 to launch the extension development host

Installing the Extension

Package the extension:

npm run package  # Build with webpack
npx vsce package # Create VSIX

Install the generated .vsix file in VSCode:
- Open VSCode
- Go to Extensions view (Ctrl+Shift+X)
- Click the ... menu and select "Install from VSIX..."
- Select the generated .vsix file

Usage

First Time Setup

On first launch, Vivus will automatically ask if you'd like to set up chat history storage. This is completely optional but recommended for preserving your conversations.

Opening the Interface

Activity Bar: Click the 🤖 robot icon on the left sidebar
Keyboard Shortcut: Ctrl+Shift+V (or Cmd+Shift+V on Mac)
Command Palette: Ctrl+Shift+P → "Open Vivus Chat"
Context Menu: Right-click selected code → "Ask about selected code"

Interface Overview

The sidebar contains three sections:

Chat: Interactive conversation with the AI assistant
Output: Detailed logs of tool executions and API calls
History: Browse and restore previous chat sessions (auto-setup available)

Available Tools

The agent has access to several tools to help with your coding tasks:

File System Tools

read_file: Read file contents with optional line ranges
write_file: Create or overwrite files
edit_file: Replace text within files
list_directory: List files and directories
create_directory: Create directories

Terminal Tools

run_command: Execute shell commands in the workspace
check_command_exists: Verify if commands are available

Search Tools

search_files: Search for text patterns within files
find_files: Find files by name patterns
get_file_info: Get detailed file information

Example Interactions

User: "Create a new TypeScript file called utils.ts with some helper functions"
Agent: [Uses write_file tool to create the file with TypeScript helper functions]

User: "Find all TODO comments in my project"
Agent: [Uses search_files tool to find TODO patterns across the codebase]

User: "Run the build command and fix any errors"
Agent: [Uses run_command to execute build, then uses file tools to fix identified issues]

Configuration

Basic Settings

Configure the extension in VSCode settings or click the ⚙️ Settings button in the chat panel:

{
  "vivusAgent.apiUrl": "https://llm.changeme.ai",
  "vivusAgent.model": "qwen3-4b-thinking",
  "vivusAgent.temperature": 0.7,
  "vivusAgent.maxTokens": 2048,
  "vivusAgent.enableStreaming": true,
  "vivusAgent.timeoutMs": 300000,
  "vivusAgent.retryAttempts": 3
}

⏱️ Timeout Settings for Thinking Models:

Default timeout increased to 5 minutes (300000ms) for qwen3-4b-thinking
Thinking models need more time to process complex requests
Adjust vivusAgent.timeoutMs if you need longer/shorter timeouts

History Setup (Optional)

Easy 2-Step Setup: The extension will automatically offer to set up chat history on first use.

Manual Setup:

Click the 🗄️ Setup History button in the History panel, OR
Use Command Palette: Ctrl+Shift+P → "Vivus: Setup Chat History"
Enter your Appwrite server URL (e.g., https://your-appwrite.com/v1)
Enter your Appwrite Project ID
Done! ✅ Database, collection, and schema are created automatically

What gets configured automatically:

✅ Tests connection to your Appwrite server
✅ Configures all VSCode settings automatically
✅ Enables history storage
✅ Sets default database (vivus-agent-history) and collection (chat-sessions) names

Requirements:

An Appwrite server (free hosting available at Appwrite Cloud)
A Project ID from your Appwrite console
Manual Step: Create the database and collection in Appwrite console (or the extension will show helpful error messages)

Database Schema (create manually in Appwrite console):

Database ID: vivus-agent-history
Collection ID: chat-sessions
Attributes: title (string), messages (string), createdAt (datetime), updatedAt (datetime), userId (string, optional)

Settings

vivusAgent.apiUrl: Local AI API URL (default: "https://llm.changeme.ai")
vivusAgent.model: Model name to use (default: "qwen3-4b-thinking")
vivusAgent.temperature: Response randomness (0-2, default: 0.7)
vivusAgent.maxTokens: Maximum tokens for responses (default: 2048)

Development

Project Structure

src/
├── extension.ts          # Extension entry point
├── chatPanel.ts         # Chat UI webview
├── lmStudioClient.ts    # Local AI API client
├── toolManager.ts       # Tool execution manager
└── tools/
    ├── fileSystemTool.ts  # File operations
    ├── terminalTool.ts    # Shell commands
    └── searchTool.ts      # Search functionality

Building

# Compile TypeScript
npm run compile

# Watch mode for development
npm run watch

Testing

The extension includes comprehensive tool implementations that mirror Amp's capabilities:

File Operations: Full CRUD operations on files and directories
Terminal Integration: Safe command execution with timeout controls
Search Functionality: Efficient text and file search with filtering
Error Handling: Robust error handling and user feedback

Contributing

Fork the repository
Create a feature branch
Make your changes
Test thoroughly
Submit a pull request

License

MIT License - see LICENSE file for details

Troubleshooting

Common Issues

Connection Failed: Verify Local AI is running at the configured URL
Tool Errors: Check workspace permissions and file paths
Model Not Found: Ensure Vivus name matches your Local AI configuration

Debugging

Enable developer tools in the extension host:

Press Ctrl+Shift+P in the extension development host
Run "Developer: Toggle Developer Tools"
Check the Console tab for error messages

Roadmap

[ ] Streaming responses for better UX
[ ] More advanced code analysis tools
[ ] Integration with VSCode's language services
[ ] Custom tool plugins
[ ] Better error recovery and retries
[ ] Conversation persistence
[ ] Multiple model support

✅ LocalAI Advanced Features Integration (v0.5.814)

Completed Features:

Timing Information: Added enableTiming setting to capture inference timing via Extra-Usage header
Parallel Requests: Added parallelRequests setting for concurrent LocalAI requests (requires LocalAI PARALLEL_REQUESTS=true)
Enhanced Logging: Detailed timing breakdowns for prompt processing and token generation
Configuration Management: New VS Code settings for LocalAI-specific optimizations

Configuration Options:

{
  "vivusAgent.enableTiming": true,     // Enable timing information
  "vivusAgent.parallelRequests": true  // Enable parallel requests
}

Usage:

Set enableTiming: true to see inference timing in the developer console
Set parallelRequests: true when LocalAI has PARALLEL_REQUESTS=true configured
Monitor timing logs for performance optimization insights

Note: This extension requires a properly configured Local AI instance with tool calling support. Make sure your model supports function calling for the best experience.

Vivus Code

Vivus AI