Kama — AI Prompt Optimizer

Turn rough ideas into world-class AI prompts. 100% local. Zero config. Your data never leaves your machine.

What is Kama?

Kama is a VS Code extension that transforms your rough ideas into structured, AI-optimized prompts — tailored for Claude, GPT, Gemini, Grok, Codex, and o3.

Type something like "add OAuth login with Google" and get a detailed, context-aware prompt that knows your project's tech stack, file structure, and active code.

No API keys. No cloud. No data leaves your computer.

⚡ Quick Start

Prerequisites

• Python 3.10+ — Download (check "Add to PATH")

That's it. No build tools, no Docker, no external services.

Install

1. Install Kama from the VS Code Marketplace
2. Open a project folder
3. Click the Kama icon in the sidebar
4. Start typing your ideas

Kama auto-starts the Brain server, installs dependencies, and is ready in seconds.

How It Works

Your Idea → Security Audit → Context Scan → Prompt Engine → Quality Score → Your AI Agent

1. You type a rough idea in the Kama sidebar
2. Security Auditor checks for injection risks and credential leaks
3. Context Scanner reads your project structure, languages, and tech stack
4. Active File context from your editor is included automatically
5. Prompt Engine generates a detailed, AI-family-optimized prompt
6. Quality Scorer grades the output (A+ to D)
7. One click sends it to your AI agent (Cursor, Copilot, Claude, etc.)

Features

Configuration

Local AI Models (Optional)

Kama works out of the box using template-based prompt generation — no extra downloads needed.

For higher-quality AI-powered generation, install a local GGUF model from the sidebar:

Models download to ~/.kama/models/. A C/C++ compiler is required for local model inference (Visual Studio Build Tools on Windows, gcc/clang on Linux/macOS).

Privacy & Security

• Zero cloud dependency — All processing runs locally on your machine
• No telemetry — We don't collect any usage data
• No API keys — Everything works offline
• CORS restricted — Brain API only accepts localhost requests
• Input sanitization — Injection and credential leak detection on every request
• SHA-256 verification — Downloaded models are integrity-checked
• Rate limiting — Per-client limits on all endpoints
• Body size limits — Requests over 1 MB are rejected

Architecture

kama/
├── brain/                        # Python backend (FastAPI)
│   ├── sslm_engine.py            # Server + endpoints
│   ├── llm_backend.py            # GGUF model management
│   ├── prompt_optimizer.py       # AI-specific prompt builder
│   ├── prompt_knowledge_base.py  # Community prompt patterns
│   ├── context_scanner.py        # Project analysis
│   ├── security_auditor.py       # Security gate
│   ├── hardware_profiler.py      # Hardware-aware model recommendations
│   └── config.py                 # Environment configuration
└── extension/                    # VS Code extension (TypeScript)
    └── src/
        ├── extension.ts               # Entry point + auto-start
        ├── providers/SidebarProvider.ts # Webview UI + streaming
        ├── services/BrainClient.ts     # HTTP/SSE client
        └── utils/config.ts            # Settings

Troubleshooting

Contributing

1. Fork the repo
2. git clone and open extension/ in VS Code
3. npm install in the extension/ directory
4. Press F5 to launch Extension Development Host
5. Submit a PR

Support

If Kama saves you time, consider supporting the project:

License

MIT — Made by Ahmet Kayra Kama