Vajra - AI Coding Assistant

Enterprise-grade multi-provider AI coding assistant for VS Code

🎯 Overview

Vajra brings the power of multiple AI providers directly into your VS Code workflow. Whether you prefer local privacy with Ollama, cutting-edge models from OpenAI and Anthropic, or specialized coding models from Qwen and DeepSeek, Vajra has you covered.

Built as a true Cursor alternative with enterprise-grade features, local model support, and intelligent model routing.

✨ Features

🤖 Multi-Provider Support

10+ AI Providers: OpenAI, Anthropic, Qwen, DeepSeek, Mistral, Gemini, Groq, Ollama, OpenRouter, HuggingFace
Smart Model Selection: Automatically choose the best model for each task
Seamless Switching: Change providers on the fly without restarting
Cost Tracking: Monitor API usage (optional)

💻 Code Intelligence

Explain Code: Get clear explanations of complex code
Refactor: Improve code quality and readability
Debug: AI-powered debugging assistance
Optimize: Performance improvements and best practices
Generate Tests: Automatic unit test generation
Add Comments: Comprehensive code documentation
Code Generation: Create code from natural language descriptions

🏠 Local-First Privacy

Ollama Integration: Run models locally on your machine
No Data Sharing: Your code stays on your computer
Offline Capable: Works without internet when using local models
Zero API Costs: Use local models completely free

🎨 Developer Experience

Context-Aware: Understands your code and project
Inline Suggestions: Right-click context menu integration
Chat Interface: Interactive AI assistant panel
Model Status: Real-time model availability and recommendations
Hardware Detection: Automatically recommends best models for your system

📦 Installation

Open VS Code
Go to Extensions (Ctrl+Shift+X / Cmd+Shift+X)
Search for "Vajra"
Click Install

That's it! You're ready to go.

🚀 Quick Start

Option 1: Local with Ollama (Recommended - Free & Private)

Why Ollama?

✅ 100% Free - No API costs ever
🔒 Private - Your code never leaves your machine
🚀 Fast - No network latency
📦 No Rate Limits - Use as much as you want

Setup (5 minutes):

Install Ollama

# macOS/Linux
curl -fsSL https://ollama.ai/install.sh | sh

# Windows
# Download from https://ollama.ai/download

Install a Coding Model

# Recommended: Best coding performance (4.1GB)
ollama pull qwen2.5-coder:7b

# Or lightweight option (1.2GB)
ollama pull qwen2.5-coder:1.5b

Configure Vajra
- Press Ctrl+Shift+P / Cmd+Shift+P
- Run: "Vajra: Select AI Provider"
- Choose: "Ollama (Local)"
- Done! ✅

Option 2: Cloud Providers (OpenAI, Anthropic, etc.)

Why Cloud?

🎯 Latest Models - Access to GPT-4, Claude 3.5, etc.
💪 Powerful - State-of-the-art AI capabilities
☁️ No Local Resources - Runs entirely in the cloud

Setup (2 minutes):

Get an API Key:
- OpenAI - GPT-4 Turbo, GPT-4o
- Anthropic - Claude 3.5 Sonnet
- Groq - Ultra-fast inference (free tier)
Configure Vajra:
- Press Ctrl+Shift+P / Cmd+Shift+P
- Run: "Vajra: Select AI Provider"
- Choose your provider
- Enter API key when prompted
- Done! ✅

🎮 How to Use

Chat Interface

Open Vajra Chat:
- Click the Vajra icon in the sidebar
- Or press Ctrl+Shift+P and run "Vajra: Open Chat"

Ask Questions:

How do I implement a binary search in Python?

Get Help with Code:
- Select code in your editor
- Ask your question in chat
- Vajra automatically includes the selected code as context

Select any code, then right-click to access:

Action	Description
Explain Selected Code	Get detailed explanations of what the code does
Refactor Selected Code	Improve code quality and readability
Debug Code with AI	Find and fix bugs with AI assistance
Optimize Code Performance	Get performance improvements
Add Comments to Code	Auto-generate comprehensive comments
Generate Unit Tests	Create test cases for your code

All Available Commands

Press Ctrl+Shift+P / Cmd+Shift+P and type "Vajra" to see:

Core Commands:

Vajra: Open Chat - Open the interactive chat panel
Vajra: Select AI Provider - Switch between providers (OpenAI, Claude, Ollama, etc.)
Vajra: Select AI Model - Change the model within your current provider

Model Management:

Vajra: Show Model Status & Recommendations - See installed models and performance benchmarks
Vajra: Check Model Availability - Check if a specific model is installed
Vajra: Get Setup Recommendations - Get personalized setup guide for your hardware

⚙️ Configuration

Quick Settings

Press Ctrl+, / Cmd+, to open Settings
Search for "Vajra"
Configure your preferences

For Ollama (Local - Free)

{
  "vajra.defaultProvider": "ollama",
  "vajra.defaultModel": "qwen2.5-coder:7b",
  "vajra.temperature": 0.7
}

For OpenAI (Cloud)

{
  "vajra.defaultProvider": "openai",
  "vajra.defaultModel": "gpt-4-turbo",
  "vajra.openaiApiKey": "sk-your-key-here",
  "vajra.temperature": 0.7
}

For Anthropic Claude (Cloud)

{
  "vajra.defaultProvider": "anthropic",
  "vajra.defaultModel": "claude-3-5-sonnet-20241022",
  "vajra.anthropicApiKey": "sk-ant-your-key-here",
  "vajra.temperature": 0.7
}

All Settings Reference

Setting	Description	Default
`vajra.defaultProvider`	Which AI provider to use	`ollama`
`vajra.defaultModel`	Which model to use	`qwen2.5-coder:7b`
`vajra.temperature`	Creativity level (0-2)	`0.7`
`vajra.maxTokens`	Maximum response length	`4096`
`vajra.ollamaEndpoint`	Ollama server URL	`http://localhost:11434`

API Keys (for cloud providers):

vajra.openaiApiKey - For GPT-4 models
vajra.anthropicApiKey - For Claude models
vajra.qwenApiKey - For Qwen models
vajra.deepseekApiKey - For DeepSeek models
vajra.mistralApiKey - For Mistral models
vajra.geminiApiKey - For Gemini models
vajra.groqApiKey - For Groq inference
vajra.openrouterApiKey - For OpenRouter access
vajra.huggingfaceApiKey - For HuggingFace models

💡 Tip: Store API keys in User Settings (Global), not workspace settings, to avoid committing secrets to git!

🎯 Supported Providers

Cloud Providers

Provider	Models	Best For	Cost
OpenAI	GPT-4 Turbo, GPT-4o, GPT-4o-mini	General purpose, Fast	$$$
Anthropic	Claude 3.5 Sonnet/Haiku	Reasoning, Large context	$$$
Qwen	Qwen2.5-Coder (32B/14B/7B/1.5B)	Coding excellence	$$
DeepSeek	DeepSeek-Coder V2.5	Cost-effective coding	$
Mistral	Codestral	Fast coding	$$
Gemini	Gemini 2.0 Flash	Multimodal	$$
Groq	Llama 3.3 70B	Ultra-fast inference	$

Local Models (Ollama - Free!)

Model	Size	Performance	Best For	Memory
qwen2.5-coder:32b	18.9GB	🏆 Highest (94.3% MBPP)	Maximum accuracy	32GB+
qwen2.5-coder:14b	8.2GB	⭐ Excellent (91.1% MBPP)	Balanced	16GB+
qwen2.5-coder:7b	4.1GB	⭐ Great (88.4% MBPP)	Best value	8GB+
qwen2.5-coder:1.5b	1.2GB	✅ Good (75.2% MBPP)	Fast & light	4GB+

⭐ Recommended: qwen2.5-coder:7b for most users

💡 Usage Examples

Example 1: Explain Complex Code

Select the code
Right-click → "Explain Selected Code"
Get instant explanation in chat

Example 2: Generate Unit Tests

Select your function
Right-click → "Generate Unit Tests"
Get comprehensive test cases

Example 3: Debug with AI

Select problematic code
Right-click → "Debug Code with AI"
Get detailed analysis and fixes

Example 4: Switch Providers on the Fly

Press Ctrl+Shift+P
Run "Vajra: Select AI Provider"
Choose a different provider for different tasks:
- Ollama for quick questions (free, fast)
- Claude for complex reasoning (best quality)
- Groq for ultra-fast responses

🔧 Troubleshooting

"Model does not exist" Error

Problem: Provider and model are mismatched (e.g., trying to use an Ollama model with OpenAI)

Solution:

Press Ctrl+Shift+P
Run "Vajra: Select AI Provider"
Choose your desired provider
Vajra will automatically select the correct model ✅

Ollama Not Detected

Check if running:

ollama list

Start Ollama:

ollama serve

Install a model:

ollama pull qwen2.5-coder:7b

API Key Issues

Verify the key is correct (no extra spaces)
Check billing is set up on provider's website
Test your key on provider's dashboard
Check rate limits - you might have hit your quota

Slow Performance

Solutions:

Use a smaller model (qwen2.5-coder:1.5b)
Switch to Groq for ultra-fast cloud inference
Use Ollama locally to eliminate network latency
Reduce maxTokens in settings

"No models available" with Ollama

Install a model:

ollama pull qwen2.5-coder:7b

Check status:

Press Ctrl+Shift+P
Run "Vajra: Show Model Status & Recommendations"

🎓 Best Practices

For Privacy

✅ Use Ollama with local models
✅ Your code never leaves your machine
✅ No API keys needed
✅ No usage tracking

For Performance

Quick tasks: Use qwen2.5-coder:7b or smaller
Complex tasks: Use qwen2.5-coder:32b or Claude 3.5
Ultra-fast: Use Groq for instant responses

For Cost

Free: Ollama (100% free forever)
Budget: DeepSeek ($1/1M tokens)
Premium: Reserve GPT-4/Claude for complex tasks only

Hardware Recommendations

Get personalized recommendations:

Press Ctrl+Shift+P
Run "Vajra: Get Setup Recommendations"

General guidelines:

4GB RAM: qwen2.5-coder:1.5b or cloud providers
8GB RAM: qwen2.5-coder:7b ⭐ (recommended)
16GB RAM: qwen2.5-coder:14b or deepseek-coder-v2:16b
32GB+ RAM: qwen2.5-coder:32b (maximum performance)

❓ FAQ

Q: Is Vajra free?
A: Yes! Use Ollama with local models for 100% free AI. Cloud providers require API keys and have usage costs.

Q: Do I need an internet connection?
A: Not if you use Ollama! Local models work completely offline.

Q: Which provider is best?
A: Depends on your needs:

Privacy: Ollama
Performance: Claude 3.5 or Qwen2.5-Coder 32B
Speed: Groq
Cost: DeepSeek or Ollama

Q: Can I use multiple providers?
A: Yes! Switch providers anytime with "Vajra: Select AI Provider"

Q: Is my code sent to the cloud?
A: Only if you use cloud providers (OpenAI, Anthropic, etc.). With Ollama, everything stays on your machine.

Q: What's the difference between this and GitHub Copilot?
A: Vajra offers:

Multiple AI providers (10+)
Local model support (Ollama)
Full chat interface
Explicit code actions
More control over models

📞 Support

Need help? We're here for you!

📧 Email: ashishjsharda@gmail.com
🐛 Report Bug: Use "Report Issue" button in VS Code
⭐ Feature Request: Use marketplace reviews or email
💬 Questions: Email us anytime

🌟 Rate Vajra

If you find Vajra helpful, please:

⭐ Rate us on the VS Code Marketplace
✍️ Write a review sharing your experience
📢 Tell your friends about Vajra

Your feedback helps us improve! 🙏

Made with ❤️ for developers everywhere

Version 0.3.2 | Email Support | Report Issue

Vajra - AI Coding Assistant

Ashish Sharda

Vajra - AI Coding Assistant

🎯 Overview

✨ Features

🤖 Multi-Provider Support

💻 Code Intelligence

🏠 Local-First Privacy

🎨 Developer Experience

📦 Installation

🚀 Quick Start

Option 1: Local with Ollama (Recommended - Free & Private)

Option 2: Cloud Providers (OpenAI, Anthropic, etc.)

🎮 How to Use

Chat Interface

Code Actions (Right-Click Menu)

All Available Commands

⚙️ Configuration

Quick Settings

For Ollama (Local - Free)

For OpenAI (Cloud)

For Anthropic Claude (Cloud)

All Settings Reference

🎯 Supported Providers

Cloud Providers

Local Models (Ollama - Free!)

💡 Usage Examples

Example 1: Explain Complex Code

Example 2: Generate Unit Tests

Example 3: Debug with AI

Example 4: Switch Providers on the Fly

🔧 Troubleshooting

"Model does not exist" Error

Ollama Not Detected

API Key Issues

Slow Performance

"No models available" with Ollama

🎓 Best Practices

For Privacy

For Performance

For Cost

Hardware Recommendations

❓ FAQ

📞 Support

🌟 Rate Vajra