cpp-boilerplate

A collection of C++ boilerplates that make systems-level coding feel as clean as Python. Each boilerplate is a standalone starting point for a specific domain — HTTP clients, ML inference, CUDA, and servers.

---

Philosophy

C++ gives you full control but at the cost of boilerplate. This repo fixes that by giving you pre-wired, production-aware starting points so you spend time on logic, not setup.

---

Boilerplates

✅ HTTP Client (/http-client)

Status: Done

Sync and async HTTP client built on libcurl + nlohmann/json. Feels like using httpx in Python.

Features

• Sync GET / POST via Client
• Async GET / POST via AsyncClient (std::async)
• Semaphore for concurrency control
• JSON body serialization via nlohmann/json
• Header and timeout support per request

Dependencies

• libcurl
• nlohmann/json

Quick start

std::vector<std::string> headers = {
    "Content-Type: application/json",
    "Authorization: Bearer my_token"
};

// Sync GET
Data req("https://api.example.com/data", headers, 10);
std::string res = Client::get(req);

// Async POST
nlohmann::json payload = {{"prompt", "hello"}, {"model", "gpt-4"}};
Data req2("https://api.example.com/chat", headers, payload, 30);
auto future = AsyncClient().post(req2);
std::string res2 = future.get();

---

🔲 ML Models (/ml)

Status: Planned

Run inference on ONNX / TensorFlow Lite / llama.cpp models with a clean API. Feels like using transformers or torch in Python.

Planned features

• Model loader abstraction (ONNX Runtime, llama.cpp)
• Tokenizer helpers
• Batched inference
• Async inference calls

Planned dependencies

• onnxruntime
• llama.cpp

---

🔲 CUDA / GPU (/cuda)

Status: Planned

GPU compute boilerplate with memory management helpers and kernel launcher wrappers. Feels like writing numba or cupy but in C++.

Planned features

• Device/host memory RAII wrappers
• Kernel launch helpers with error checking
• cuBLAS / cuDNN setup
• Multi-GPU support

Planned dependencies

• CUDA Toolkit
• cuBLAS
• cuDNN

---

🔲 Server (/server)

Status: Planned

HTTP server boilerplate with routing, middleware, and JSON request/response handling. Feels like FastAPI in Python.

Planned features

• Route registration (GET, POST, PUT, DELETE)
• JSON body parsing
• Middleware chain
• Async request handling

Planned dependencies

• drogon
• nlohmann/json

---

Project Structure

cpp-boilerplate/
├── http-client/
│   ├── src/
│   │   └── main.cpp
│   ├── lib/
│   │   ├── client.cpp
│   │   └── client.h
│   ├── include/
│   └── CMakeLists.txt
├── ml/
├── cuda/
├── server/
└── README.md

---

Building (per boilerplate)

cd http-client
cmake -B build
cmake --build build
./build/boilerplate

CMake is configured to auto-detect all .cpp files in src/ and lib/ — just drop your files in and rebuild.

---

Requirements

• C++17 or later
• CMake 3.20+
• Per-boilerplate dependencies listed above