⚡ LangShift

AI Code Transpiler for VS Code

Convert code between 25+ languages with a single rename. Six AI providers. Smart caching. PII scrubbing. Offline support. Enterprise-ready.

Install from Marketplace · Report Bug · Request Feature

  main.py   ──rename──▶   main.java
  ┌──────────────┐        ┌──────────────────────┐
  │ def greet(): │   AI   │ public String greet() │
  │   return "Hi"│  ────▶ │   { return "Hi"; }   │
  └──────────────┘        └──────────────────────┘

🤔 Why LangShift?

Most code translators are toys — copy-paste into a web UI, get broken output, manually fix imports. LangShift is different.

	Other Tools	LangShift v2.0
Workflow	Copy → paste → web UI → copy back	Rename the file. Done.
Providers	Single vendor lock-in	6 providers (4 cloud + 2 local)
Offline	❌ Always needs internet	✅ Ollama & LM Studio work air-gapped
Context	Converts file in isolation	Reads import tree for accurate output
Validation	Trust the AI blindly	Compiler checks + confidence scoring
Privacy	Your code hits their servers	PII scrubbing, audit logs, local options
Caching	Recomputes every time	SHA-256 smart cache — instant repeats
Enterprise	Settings per user	`.langshiftrc.json` + team shared config
Reliability	Fails? You retry.	Fallback chain auto-retries next provider

✨ Feature Highlights

🧠 AI-Powered Core

6 AI Providers — Claude, GPT-4o, Gemini, OpenRouter, Ollama, LM Studio
Two-Pass Conversion — AI generates code, then reviews and fixes its own output
Confidence Scoring — Every conversion rated 0–100 so you know what to double-check
Model Routing — Auto-selects the best model for each language pair
Few-Shot Examples — Built-in examples for popular conversion pairs boost accuracy
Streaming Output — Watch converted code appear in real-time

🔒 Enterprise Security

PII Scrubbing — Auto-redact emails, IPs, phone numbers, SSNs before sending to AI (HIPAA/GDPR safe)
Audit Log — JSONL log of every conversion with CSV export for compliance (no code content logged)
Secret Scanner — Warns before sending code containing hardcoded credentials
Encrypted Key Storage — API keys stored in VS Code SecretStorage (OS keychain), never in plaintext
Local AI Options — Ollama & LM Studio keep all data on your machine

📦 Multi-File Intelligence

Import Tree Context — Reads imported/required files for accurate type-aware conversions
Batch Folder Conversion — Right-click a folder → convert all files at once
Auto Test Detection — Finds and converts associated test files automatically
Import Resolver — Checks if target-language packages actually exist

⚡ Performance & Reliability

Smart Caching — SHA-256 content hashing; unchanged files convert instantly from cache
Provider Fallback Chain — Primary fails with 429/5xx? Auto-retries the next provider
Compiler Validation — Runs tsc, javac, go build, rustc, g++, gcc, dotnet build, python -m py_compile on output
Undo with Backup — Every file backed up atomically; one-click restore

🛠 Team & Configuration

Custom Rules — .langshiftrc.json per-project overrides (model, language mappings, exclusions)
Team Settings — .vscode/langshift.team.json shared via version control
Dashboard — Visual analytics: conversion history, provider breakdown, cache hit rate
Proxy Support — HTTP/HTTPS proxy for corporate networks

🌍 Supported Languages


Python	JavaScript	TypeScript	Java	C++
C	C#	Go	Rust	Ruby
PHP	Swift	Kotlin	Scala	Dart
Lua	Perl	Haskell	Elixir	R
MATLAB	Shell/Bash	PowerShell	JSX	TSX

25 languages with full extension detection and idiomatic conversion patterns.

🚀 Quick Start

1️⃣ Install

Search "LangShift" in the VS Code Extensions panel, or run:

ext install langshift.langshift

2️⃣ Configure an AI Provider

Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P) → LangShift: Configure API Key

Select your provider → paste your API key → Done.

💡 No API key needed for Ollama or LM Studio — they run locally on your machine.

3️⃣ Convert Code

Option A — Rename to convert (magic mode):

app.py  →  app.java         # Python → Java  
server.js  →  server.ts     # JavaScript → TypeScript  
utils.go  →  utils.rs       # Go → Rust

Option B — Command Palette: Ctrl+Shift+P → LangShift: Convert This File → select target language

Option C — Right-click: Right-click any file in Explorer → LangShift: Convert This File

A confirmation dialog appears → choose Convert, Preview Diff, or Skip.

🔧 Provider Setup

☁️ Cloud Providers

🟣 Anthropic (Claude) — Recommended for code quality

Get an API key at console.anthropic.com
Ctrl+Shift+P → LangShift: Configure API Key → select Anthropic
Paste your key

Available models: | Model | Best For | |---|---| | claude-sonnet-4-20250514 (default) | Best balance of quality and speed | | claude-haiku-3-5 | Fast, cost-effective conversions |

"langshift.aiProvider": "anthropic",
"langshift.anthropicModel": "claude-sonnet-4-20250514"

🟢 OpenAI (GPT-4o) — Strong all-rounder

Get an API key at platform.openai.com/api-keys
Ctrl+Shift+P → LangShift: Configure API Key → select OpenAI
Paste your key

Available models: | Model | Best For | |---|---| | gpt-4o (default) | Highest accuracy | | gpt-4o-mini | Budget-friendly, still strong | | o1-preview | Complex architectural conversions | | o1-mini | Reasoning-heavy, smaller footprint |

"langshift.aiProvider": "openai",
"langshift.openaiModel": "gpt-4o"

🔵 Google Gemini — Fast and cost-effective

Get an API key at aistudio.google.com/app/apikey
Ctrl+Shift+P → LangShift: Configure API Key → select Gemini
Paste your key

Available models: | Model | Best For | |---|---| | gemini-2.0-flash (default) | Ultra-fast with great quality | | gemini-1.5-pro | Complex, large-file conversions | | gemini-1.5-flash | Budget-friendly speed |

"langshift.aiProvider": "gemini",
"langshift.geminiModel": "gemini-2.0-flash"

🟡 OpenRouter — Access 200+ models with one key

Get an API key at openrouter.ai/keys
Ctrl+Shift+P → LangShift: Configure API Key → select OpenRouter
Paste your key

Example models:

"langshift.aiProvider": "openrouter",
"langshift.openrouterModel": "anthropic/claude-sonnet-4-20250514"

Other popular choices:

meta-llama/llama-3.3-70b-instruct
google/gemini-2.0-flash-exp
mistralai/mixtral-8x7b-instruct

🏠 Local Providers (No API Key Required)

🦙 Ollama — Fully offline, air-gapped safe

Install Ollama from ollama.com
Pull a code model:
```
ollama pull deepseek-coder-v2
```

Set your VS Code settings:

"langshift.aiProvider": "ollama",
"langshift.ollamaModel": "deepseek-coder-v2",
"langshift.ollamaBaseUrl": "http://localhost:11434/v1"

No API key. No data leaves your machine. Perfect for classified/regulated environments.

💡 Run ollama list to see available models on your system.

🖥 LM Studio — GUI-based local AI

Install LM Studio from lmstudio.ai
Download and load a code model in the LM Studio UI
Start the local server in LM Studio

Set your VS Code settings:

"langshift.aiProvider": "lmstudio",
"langshift.lmstudioModel": "loaded-model",
"langshift.lmstudioBaseUrl": "http://localhost:1234/v1"

No API key. No data leaves your machine. Use any GGUF model you like.

⚙️ Configuration Reference

Open VS Code Settings (Ctrl+,) → search "langshift", or edit settings.json directly.

Provider Settings

Setting	Type	Default	Description
`langshift.aiProvider`	`string`	`"anthropic"`	Active provider: `anthropic`, `openai`, `gemini`, `openrouter`, `ollama`, `lmstudio`
`langshift.anthropicModel`	`string`	`"claude-sonnet-4-20250514"`	Anthropic model
`langshift.openaiModel`	`string`	`"gpt-4o"`	OpenAI model
`langshift.geminiModel`	`string`	`"gemini-2.0-flash"`	Google Gemini model
`langshift.openrouterModel`	`string`	`"anthropic/claude-sonnet-4-20250514"`	OpenRouter model identifier
`langshift.ollamaModel`	`string`	`"deepseek-coder-v2"`	Ollama model name
`langshift.ollamaBaseUrl`	`string`	`"http://localhost:11434/v1"`	Ollama API endpoint
`langshift.lmstudioModel`	`string`	`"loaded-model"`	LM Studio model identifier
`langshift.lmstudioBaseUrl`	`string`	`"http://localhost:1234/v1"`	LM Studio API endpoint

Conversion Behavior

Setting	Type	Default	Description
`langshift.autoConvertOnRename`	`bool`	`true`	Trigger conversion when a file is renamed to a new extension
`langshift.showConfirmationDialog`	`bool`	`true`	Show confirm dialog before converting
`langshift.showDiffBeforeApply`	`bool`	`true`	Side-by-side diff preview before applying
`langshift.autoConvertTests`	`bool`	`true`	Auto-detect and convert associated test files
`langshift.preserveComments`	`bool`	`true`	Preserve and translate comments/docstrings
`langshift.streaming`	`bool`	`true`	Show converted code appearing in real-time
`langshift.backupBeforeConvert`	`bool`	`true`	Create backup before each conversion

AI Enhancement

Setting	Type	Default	Description
`langshift.twoPassConversion`	`bool`	`false`	AI reviews its own output in a second pass (doubles API cost)
`langshift.confidenceScoring`	`bool`	`true`	AI rates conversion confidence 0–100
`langshift.modelRouting`	`bool`	`false`	Use curated provider/model routes when the recommended provider is configured
`langshift.multiFileContext`	`bool`	`true`	Include imported files as context for better accuracy
`langshift.maxContextFiles`	`number`	`5`	Max number of context files to include
`langshift.maxContextLinesPerFile`	`number`	`200`	Max lines per context file
`langshift.fallbackEnabled`	`bool`	`false`	Auto-retry with fallback providers on 429/5xx errors
`langshift.fallbackChain`	`array`	`["openai","gemini"]`	Ordered list of fallback providers

Security & Compliance

Setting	Type	Default	Description
`langshift.piiScrubbing`	`bool`	`false`	Auto-redact emails, IPs, phones, SSNs before sending to AI
`langshift.auditLog`	`bool`	`false`	Log conversions to `.langshift/audit.jsonl` (no code content)
`langshift.enableCache`	`bool`	`true`	SHA-256 smart caching for instant repeat conversions
`langshift.proxyUrl`	`string`	`""`	HTTP/HTTPS proxy for corporate networks
`langshift.telemetryEnabled`	`bool`	`false`	Opt-in anonymous telemetry (no code ever sent)

Limits

Setting	Type	Default	Description
`langshift.maxFileSizeKB`	`number`	`500`	Maximum file size (KB) for conversion
`langshift.conversionTimeout`	`number`	`90`	Timeout in seconds per AI request
`langshift.rateLimitPerHour`	`number`	`50`	Maximum conversions per hour
`langshift.batchDelayMs`	`number`	`600`	Delay (ms) between batch conversion requests

📄 `.langshiftrc.json` — Per-Project Rules

Create a .langshiftrc.json in your project root to override settings per-project:

{
  "defaultProvider": "anthropic",
  "defaultModel": "claude-sonnet-4-20250514",
  "twoPassConversion": true,
  "confidenceThreshold": 75,
  "rules": [
    {
      "from": "python",
      "to": "java",
      "model": "claude-sonnet-4-20250514",
      "instructions": "Use Java 17+ features. Prefer records over POJOs. Use var where type is obvious."
    },
    {
      "from": "javascript",
      "to": "typescript",
      "model": "gpt-4o",
      "instructions": "Use strict mode. Prefer interfaces over type aliases. Add JSDoc where missing."
    }
  ],
  "exclude": [
    "**/node_modules/**",
    "**/vendor/**",
    "**/*.generated.*"
  ],
  "piiScrubbing": true,
  "auditLog": true
}

💡 The rules array lets you set per-language-pair models and custom instructions — ideal for teams with specific coding standards.

👥 Team Settings — `.vscode/langshift.team.json`

Share team-wide LangShift configuration through version control:

{
  "version": "2.0",
  "provider": "anthropic",
  "model": "claude-sonnet-4-20250514",
  "twoPassConversion": true,
  "confidenceScoring": true,
  "piiScrubbing": true,
  "auditLog": true,
  "fallbackEnabled": true,
  "fallbackChain": ["openai", "gemini"],
  "maxFileSizeKB": 300,
  "rateLimitPerHour": 30,
  "conversionTimeout": 120,
  "exclude": [
    "**/generated/**",
    "**/vendor/**"
  ],
  "rules": [
    {
      "from": "python",
      "to": "java",
      "instructions": "Follow our Java 17 style guide. Use Spring Boot annotations."
    }
  ]
}

Commit this file to your repo. Every team member gets the same conversion behavior — no per-user configuration drift.

📋 Commands

Command	Shortcut	Description
LangShift: Convert This File	—	Convert the active file to another language
LangShift: Convert All Files in Folder	—	Batch convert an entire folder
LangShift: Show Conversion History	—	Browse past conversions with details
LangShift: Open Dashboard	—	Visual analytics dashboard
LangShift: Configure API Key	—	Add or change a provider's API key
LangShift: Delete Stored API Key	—	Remove a stored API key
LangShift: Undo Last Conversion	—	Restore previous file from backup
LangShift: Cancel Active Conversion	—	Cancel an in-flight AI request
LangShift: Clear Conversion Cache	—	Clear the SHA-256 smart cache
LangShift: Export Audit Log (CSV)	—	Export audit log as CSV for compliance

All commands are also accessible via the right-click context menu on supported files and folders.

🏗 Architecture Overview

src/
├── extension.ts                  ← Activation, command registration, rename watcher
│
├── ai/                           ← Provider abstraction layer
│   ├── AIProvider.ts             ← Common interface for all providers
│   ├── AnthropicProvider.ts      ← Claude API integration
│   ├── OpenAIProvider.ts         ← GPT-4o API integration
│   ├── GeminiProvider.ts         ← Gemini API integration
│   ├── OpenRouterProvider.ts     ← OpenRouter multi-model gateway
│   ├── OllamaProvider.ts         ← Local Ollama integration
│   ├── LMStudioProvider.ts       ← Local LM Studio integration
│   └── ProviderFactory.ts        ← Factory pattern for provider instantiation
│
├── core/                         ← Conversion engine & intelligence
│   ├── ConversionEngine.ts       ← Orchestrates the full conversion pipeline
│   ├── LanguageDetector.ts       ← 25 languages, extension ↔ name mapping
│   ├── ConversionHistory.ts      ← Persistent history with stats
│   ├── ConversionCache.ts        ← SHA-256 content-hash smart cache
│   ├── BackupManager.ts          ← Atomic backup and restore
│   ├── CodeValidator.ts          ← Syntax validation per language
│   ├── CompilerValidator.ts      ← Runs actual compilers on output
│   ├── ContextCollector.ts       ← Import tree walker for multi-file context
│   ├── FallbackChain.ts          ← Provider fallback with retry logic
│   ├── FewShotExamples.ts        ← Built-in examples for popular pairs
│   ├── ImportResolver.ts         ← Validates target-language package existence
│   ├── ModelRouter.ts            ← Auto-picks best model per language pair
│   ├── PIIRedactor.ts            ← Regex-based PII scrubbing engine
│   └── AuditLog.ts               ← JSONL audit trail with CSV export
│
├── config/                       ← Configuration management
│   ├── CustomRules.ts            ← .langshiftrc.json loader and validator
│   └── TeamSettings.ts           ← Team shared settings loader
│
├── security/                     ← Security layer
│   ├── SecurityManager.ts        ← Multi-provider encrypted key storage + secret scanning
│   └── RateLimiter.ts            ← Sliding-window rate limiter
│
├── providers/                    ← VS Code UI providers
│   ├── DiffProvider.ts           ← Side-by-side diff editor + dashboard webview
│   └── SidebarProvider.ts        ← Activity bar sidebar with live stats
│
└── utils/                        ← Shared utilities
    ├── ConfigManager.ts          ← Typed settings wrapper
    ├── Logger.ts                 ← VS Code output channel logger
    └── TelemetryService.ts       ← Privacy-first opt-in telemetry

Conversion Pipeline

┌─────────┐    ┌──────────┐    ┌───────────┐    ┌──────────┐    ┌───────────┐
│  Input   │───▶│  Cache   │───▶│ PII Scrub │───▶│ Context  │───▶│  AI Call  │
│  File    │    │  Check   │    │ (opt-in)  │    │ Collect  │    │ (stream)  │
└─────────┘    └──────────┘    └───────────┘    └──────────┘    └───────────┘
                 │ hit                                               │
                 ▼                                                   ▼
              ┌──────────┐                                    ┌───────────┐
              │ Instant  │                                    │ Two-Pass  │
              │ Return   │                                    │ Review    │
              └──────────┘                                    └───────────┘
                                                                    │
                ┌───────────┐    ┌──────────┐    ┌──────────┐       │
                │  Backup   │◀──│ Compiler │◀──│ Validate │◀──────┘
                │  + Apply  │    │  Check   │    │ + Score  │
                └───────────┘    └──────────┘    └──────────┘

🔐 Security & Privacy

LangShift is built with enterprise security requirements in mind:

Concern	How LangShift Handles It
API Key Storage	Encrypted via VS Code `SecretStorage` (OS keychain). Never stored in `settings.json` or plaintext files.
Code Transmission	HTTPS only. Code sent only to your chosen provider. Local providers (Ollama/LM Studio) send nothing over the network.
PII Protection	Optional regex-based scrubbing of emails, IPs, phone numbers, SSNs before AI transmission. HIPAA/GDPR-compatible.
Audit Trail	Optional JSONL logging of every conversion (timestamp, provider, languages, file name — no code content). Exportable as CSV.
Secret Scanning	Warns if source code contains hardcoded passwords, API keys, or tokens before sending.
Local Code Retention	Recent conversion history keeps source and converted code locally in VS Code global state for diff viewing; older entries drop code content. Audit logs and telemetry never include code.
Backups	Stored in `.langshift-backups/` with auto-generated `.gitignore` to prevent accidental commits.
Telemetry	Strictly opt-in. Anonymous. Never includes code, file paths, or API keys.
Air-Gapped Environments	Use Ollama or LM Studio — zero network requests, full functionality.
Corporate Proxy	Set `langshift.proxyUrl` for environments behind HTTP/HTTPS proxies.

❓ FAQ

Which AI provider should I use?

Best quality: Anthropic Claude (claude-sonnet-4-20250514) — excellent at understanding code semantics and idioms.
Best value: Google Gemini (gemini-2.0-flash) — fast, cheap, and surprisingly capable.
Most models: OpenRouter — access 200+ models with a single API key.
Offline/private: Ollama or LM Studio — no data leaves your machine.

Does my code get stored anywhere?

No. LangShift sends your code to the AI provider you choose (via HTTPS) and receives the converted output. Nothing is stored on any intermediate server. With Ollama or LM Studio, nothing leaves your machine at all.

How does smart caching work?

LangShift hashes your file content using SHA-256, combined with the source language, target language, and model. If you convert the same unchanged file with the same settings again, the result is returned instantly from cache — no API call, no cost.

What is two-pass conversion?

When enabled, LangShift sends the AI's first output back to the AI with the prompt "review and fix this conversion." The AI catches its own mistakes — missing edge cases, incorrect imports, non-idiomatic patterns. This doubles API cost but meaningfully improves output quality for complex files.

Can I use this in an air-gapped environment?

Yes. Configure Ollama or LM Studio as your provider. Both run entirely locally. No internet connection required. No telemetry. No external API calls. Perfect for classified, regulated, or restricted environments.

What does the confidence score mean?

The AI rates its own conversion confidence from 0 to 100:

90–100: High confidence — straightforward conversion, likely correct.
70–89: Good — may need minor review for edge cases.
50–69: Moderate — review recommended, especially complex logic.
Below 50: Low — significant manual review needed.

How does the fallback chain work?

When fallbackEnabled is true and your primary provider returns a 429 (rate limit) or 5xx (server error), LangShift automatically retries with the next provider in your fallbackChain. For example, with the default chain ["openai", "gemini"], if Anthropic fails, it tries OpenAI, then Gemini.

What compilers are supported for validation?

LangShift can optionally run the actual compiler/interpreter on the converted output:

Language	Compiler/Tool
TypeScript	`tsc --noEmit`
Java	`javac`
Go	`go build`
Rust	`rustc --edition 2021`
C++	`g++`
C	`gcc`
C#	`dotnet build`
Python	`python -m py_compile`

The compiler must be installed on your system. This is optional validation — conversions still work without it.

What is PII scrubbing and when should I enable it?

PII scrubbing uses regex patterns to redact personally identifiable information (emails, IP addresses, phone numbers, SSNs) from your code before it's sent to the AI. Enable it when working with code that may contain real user data — especially in healthcare (HIPAA) or European (GDPR) contexts. Redacted values are replaced with placeholders and restored after conversion.

Can my whole team share the same settings?

Yes. Create .vscode/langshift.team.json in your repo and commit it. All team members will use the same provider, model, conversion rules, and security settings. Individual user settings still override team settings where needed.

🤝 Contributing

Contributions are welcome! Here's how to get started:

# Clone the repo
git clone https://github.com/abneeshsingh21/LangShift-Extension-.git
cd langshift-vscode

# Install dependencies
npm install

# Compile
npm run compile

# Launch Extension Development Host
# Press F5 in VS Code — a new window opens with LangShift loaded

Development Commands

npm run compile    # Compile TypeScript
npm run watch      # Watch mode for development
npm test           # Run test suite
npm run lint       # Lint with ESLint
npm run package    # Package as .vsix for distribution

Guidelines

Fork the repository and create a feature branch
Write tests for new functionality
Follow the existing code style (ESLint enforced)
Update documentation for user-facing changes
Open a PR with a clear description of your changes

📖 See CONTRIBUTING.md for detailed guidelines.

📜 License

Built with ❤️ for developers who speak every language.

⬆ Back to Top

LangShift — AI Code Transpiler

abneesh singh

⚡ LangShift

AI Code Transpiler for VS Code

🤔 Why LangShift?

✨ Feature Highlights

🧠 AI-Powered Core

🔒 Enterprise Security

📦 Multi-File Intelligence

⚡ Performance & Reliability

🛠 Team & Configuration

🌍 Supported Languages

🚀 Quick Start

1️⃣ Install

2️⃣ Configure an AI Provider

3️⃣ Convert Code

🔧 Provider Setup

☁️ Cloud Providers

🏠 Local Providers (No API Key Required)

⚙️ Configuration Reference

Provider Settings

Conversion Behavior

AI Enhancement

Security & Compliance

Limits

📄 `.langshiftrc.json` — Per-Project Rules

👥 Team Settings — `.vscode/langshift.team.json`

📋 Commands

🏗 Architecture Overview

Conversion Pipeline

🔐 Security & Privacy

❓ FAQ

🤝 Contributing

Development Commands

Guidelines

📜 License

LangShift — AI Code Transpiler

abneesh singh

⚡ LangShift

AI Code Transpiler for VS Code

🤔 Why LangShift?

✨ Feature Highlights

🧠 AI-Powered Core

🔒 Enterprise Security

📦 Multi-File Intelligence

⚡ Performance & Reliability

🛠 Team & Configuration

🌍 Supported Languages

🚀 Quick Start

1️⃣ Install

2️⃣ Configure an AI Provider

3️⃣ Convert Code

🔧 Provider Setup

☁️ Cloud Providers

🏠 Local Providers (No API Key Required)

⚙️ Configuration Reference

Provider Settings

Conversion Behavior

AI Enhancement

Security & Compliance

Limits

📄 .langshiftrc.json — Per-Project Rules

👥 Team Settings — .vscode/langshift.team.json

📋 Commands

🏗 Architecture Overview

Conversion Pipeline

🔐 Security & Privacy

❓ FAQ

🤝 Contributing

Development Commands

Guidelines

📜 License

📄 `.langshiftrc.json` — Per-Project Rules

👥 Team Settings — `.vscode/langshift.team.json`