TNG.sh: Architecture & Technology Stack
TNG.sh combines modern programming languages and LLMs with proven developer tooling to create a seamless test generation experience.
How TNG.sh Works
Core Engine – Built with High-Performance Systems Programming
We experimented with C, C++, and Go before settling on Rust for its:
- ⚡ Speed – Lightning-fast parsing and analysis
- 🔒 Safety – Memory safety without garbage collection overhead
- 🔗 Strong integration – Excellent bindings with both Ruby and Python ecosystems
Code Intelligence – Advanced Static Analysis
We use advanced static code analysis with AST parsing to extract and analyze only what's necessary.
Currently supporting:
- ✅ Ruby
- ✅ Python
- 🧪 JavaScript (in beta)
- Golang
- React js
Test Generation – Fine-Tuned LLM
Our fine-tuned, pre-trained LLM specializes in test creation and understands framework patterns, generating tests that actually make sense for your codebase.
Unlike generic LLMs, our model is purpose-built for:
- Understanding framework-specific patterns (Django, Rails, FastAPI, Flask, etc.)
- Generating idiomatic tests that follow best practices
- Producing tests with proper fixtures, mocks, and assertions
Developer Interfaces – Simple & Familiar
Tools you already know and love:
- 💻 CLI packages – Terminal-based workflow for all major languages
- 🎨 VS Code extension – Generate tests with a single keyboard shortcut
- 🔄 Seamless integration – Works with your existing development workflow
Privacy by Design – Cloud-Backed, Zero Data Retention
We do not train on, sell, or keep your code.
- 🔐 Only small bits of metadata and relevant code snippets are sent to generate tests
- 🚫 Nothing is stored or reused
- ✅ Your proprietary logic stays on your machine
- 📊 Only structural information (AST metadata) is transmitted
In Short
Static analysis + LLM + developer-friendly interfaces = TNG.sh
Technical Stack at a Glance
| Component | Technology | Why |
|---|---|---|
| Core Engine | Rust | Performance, safety, ecosystem integration |
| Code Analysis | AST Parsing | Deep code understanding without execution |
| AI Model | Fine-tuned LLM | Purpose-built for test generation |
| Interfaces | CLI + VS Code | Familiar developer tools |
| Privacy | Zero retention | Your code never leaves your control |
Why This Architecture Matters
Speed: Rust-powered analysis completes in under 100ms for most files
Accuracy: AST-based parsing understands your code structurally, not just as text
Privacy: Local analysis + cloud generation with no data retention = secure workflow
Developer Experience: Works where you work—terminal or IDE
TNG.sh: Because test generation should be fast, accurate, and respectful of your privacy.
Top comments (0)