@nikita-bekish/code-assistant

🤖 AI-powered code assistant CLI with RAG (Retrieval-Augmented Generation), LLM classification, and MCP tools for frontend development.

Features

• 🔍 Semantic Search - Find relevant code using RAG with vector embeddings
• 💬 Interactive Chat - Ask questions about your codebase
• 🏷️ Smart Classification - LLM-based question routing (git/crm/tasks/rag)
• 🛠️ MCP Tools - Model Context Protocol for git operations, CRM, and task management
• 📊 Project Analysis - Automatic indexing and code understanding
• 🎯 Frontend Focus - Optimized for JavaScript/TypeScript projects

Requirements

• Node.js >= 18.0.0
• Ollama - Local LLM runtime (Installation guide)

  # Install Ollama
  curl https://ollama.ai/install.sh | sh
    
  # Pull required models
  ollama pull llama3.2
  ollama pull nomic-embed-text

Installation

Global (recommended)

npm install -g @nikita-bekish/code-assistant

Local

npm install @nikita-bekish/code-assistant
npx code-assistant --help

Quick Start

1. Initialize in your project:

cd your-project
code-assistant init

2. Index your codebase:

code-assistant index

3. Start chatting:

code-assistant chat

Commands

code-assistant init

Initialize configuration in your project.

code-assistant init

Creates .code-assistant-config.json with default settings.

code-assistant index

Index your codebase for semantic search.

code-assistant index

Options:

• Indexes all files according to .code-assistant-config.json
• Generates embeddings using Ollama
• Stores chunks in node_modules/.code-assistant/

code-assistant chat

Start interactive chat session.

code-assistant chat

Example questions:

• "How does authentication work?"
• "Show me high priority tasks"
• "What is the current git status?"
• "List all open tickets for user_1"

code-assistant reindex

Reindex the project (use after major code changes).

code-assistant reindex

Configuration

.code-assistant-config.json example:

{
  "projectName": "My Project",
  "projectDescription": "A modern web application",
  "indexing": {
    "includeFolders": ["src", "lib"],
    "excludeFolders": ["node_modules", "dist", ".git"],
    "includeFileTypes": ["js", "ts", "jsx", "tsx", "vue", "svelte"],
    "chunkSize": 1024,
    "chunkOverlap": 256
  },
  "llm": {
    "model": "llama3.2",
    "temperature": 0.7,
    "maxResults": 5
  },
  "embedding": {
    "model": "nomic-embed-text",
    "provider": "ollama"
  }
}

Features Overview

🔍 RAG (Retrieval-Augmented Generation)

• Semantic code search using vector embeddings
• Context-aware answers based on your codebase
• Conversation memory for follow-up questions

🏷️ Smart Question Classification

• Automatic routing: git → crm → tasks → rag
• LLM-based classification for ambiguous questions
• Heuristics for fast obvious cases

🛠️ MCP Tools

Git Tools:

• git_branch - Get current branch
• git_status - Show repository status

CRM Tools:

• get_user - User information
• list_tickets - User tickets
• create_ticket - New support ticket
• update_ticket - Update ticket status

Tasks Tools:

• list_tasks - Team tasks with filters
• create_task - New task
• update_task - Update task status

Examples

Basic Usage

# Initialize and index
code-assistant init
code-assistant index

# Ask about code
code-assistant chat
> How does authentication work in this project?

# Ask about tasks
> Show me high priority tasks

With OpenAI (optional)

export OPENAI_API_KEY=your_key_here
code-assistant chat

Troubleshooting

"Ollama not found"

Install Ollama from https://ollama.ai/

"Model not found"

Pull required models:

ollama pull llama3.2
ollama pull nomic-embed-text

"No index found"

Run indexing first:

code-assistant index

Slow responses

• Reduce chunkSize in config
• Use smaller LLM model
• Consider using OpenAI API

Development

# Clone repository
git clone https://github.com/nikita-bekish/my-code-assistant.git
cd my-code-assistant

# Install dependencies
npm install

# Build
npm run build

# Test locally
npm link
code-assistant --help

License

MIT © Nikita Bekish

Links

• GitHub
• Issues
• npm