Llamautoma NPM

Llamautoma

AI-powered automation server using LangChain.js, LangGraph.js, and Ollama. This server is part of the coc-llamautoma project, providing AI-assisted code generation and editing features through a streaming API.

Features

Chat: Stream-based chat interface with AI model for code-related discussions
Edit: AI-assisted code editing with file modifications and safety checks
Compose: AI-powered file generation with context awareness
Embed: File and text embedding for context-aware operations
Tools: Extensible tool system with safety checks and user confirmation
LSP Integration: Language Server Protocol support for IDE integration
Streaming: Real-time streaming responses for all operations
Safety: Built-in safety checks and user confirmation for dangerous operations
Memory: Short-term and long-term memory using LangGraph.js

Prerequisites

Bun >= 1.0.0
Ollama with required models:
- Default: qwen2.5-coder:7b (production)
- Testing: qwen2.5-coder:1.5b (faster for tests)

Installation

# Clone the repository
git clone https://github.com/dgpt/llamautoma.git
cd llamautoma

# Install dependencies
bun install

# Start the server
bun run dev

Configuration

Environment Variables

Create a .env file in the root directory:

# Server Configuration
PORT=3000                              # Server port (default: 3000)
OLLAMA_HOST=http://localhost:11434     # Ollama server URL
LOG_LEVEL=info                         # Logging level (debug|info|warn|error)

# Agent Configuration
MODEL_NAME=qwen2.5-coder:7b            # Default model name
MAX_ITERATIONS=10                      # Maximum iterations per request
USER_INPUT_TIMEOUT=30000              # Timeout for user input (ms)

Safety Configuration

The server includes built-in safety checks that can be configured:

{
  requireToolConfirmation: true,      // Require user confirmation for tool execution
  requireToolFeedback: true,          // Require user feedback after tool execution
  maxInputLength: 8192,               // Maximum input length
  dangerousToolPatterns: [            // Patterns requiring extra confirmation
    'rm -rf /',
    'DROP TABLE',
    'sudo rm',
    'wget http',
    'curl',
    'exec',
    'bash -c',
    'zsh -c',
    'sh -c'
  ]
}

API Endpoints

Chat (`POST /chat`)

Stream-based chat interface with the AI model.

// Request
{
  messages: Array<{
    role: 'user' | 'system' | 'assistant'
    content: string
  }>
  threadId: string
}

// Response (Server-Sent Events)
data: {
  messages: Array<BaseMessage>
  status: 'continue' | 'end'
  toolFeedback: Record<string, string>
  iterations: number
  threadId: string
}

Edit (`POST /edit`)

AI-assisted code editing with file modifications.

// Request
{
  file: string
  prompt: string
  threadId: string
}

// Response (Server-Sent Events)
data: {
  type: 'edit'
  messages: Array<BaseMessage>
  status: 'continue' | 'end'
  toolFeedback: Record<string, string>
  iterations: number
  threadId: string
}

Compose (`POST /compose`)

AI-powered file generation.

// Request
{
  prompt: string
  threadId: string
}

// Response (Server-Sent Events)
data: {
  type: 'compose'
  messages: Array<BaseMessage>
  status: 'continue' | 'end'
  toolFeedback: Record<string, string>
  iterations: number
  threadId: string
}

Embed (`POST /embed`)

File and text embedding for context-aware operations.

// Request
{
  type: 'file' | 'text'
  path?: string
  content: string
}

// Response
{
  success: boolean
  embeddings: number[]
}

Development

# Start development server with hot reload
bun run dev

# Run tests
bun test

# Run tests in watch mode
bun test --watch

# Run tests with coverage
bun test --coverage

# Lint code
bun run lint

# Format code
bun run format

Project Structure

llamautoma/
├── src/
│   ├── agents/         # AI agents and tools
│   ├── config/         # Configuration management
│   ├── server/         # Server implementation
│   ├── types/          # TypeScript type definitions
│   └── utils/          # Utility functions
├── tests/
│   ├── unit/          # Unit tests
│   └── integration/   # Integration tests
└── package.json