GitHub - okba14/NeuroHTTP: High-Performance AI-Native Web Server — built in C & Assembly for ultra-fast AI inference and streaming.

High-performance, AI-native server built from scratch in C + Assembly — handles heavy AI payloads with minimal latency.

🚀 Quick Start

1️⃣ AI Provider Setup (Optional)

NeuroHTTP is provider-agnostic and does not require a specific AI vendor.

You may run the server using any OpenAI-compatible API, GROQ, or even a local AI model.

If your setup requires an API key, export it as an environment variable:

export OPENAI_API_KEY="gsk_XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"

2️⃣ Install Dependencies

On Debian / Ubuntu / Kali:

sudo apt-get update
sudo apt-get install -y libcurl4-openssl-dev build-essential

3️⃣ Clone the Repository & Build the Server

git clone https://github.com/okba14/NeuroHTTP.git
cd NeuroHTTP
make rebuild

The make rebuild command compiles the server from scratch.

4️⃣ Run the Server

The server will run on port 8080 by default. Logs are displayed in the same terminal.

5️⃣ Send a Test Request (curl)

Open a second terminal and send a POST request:

curl -X POST http://localhost:8080/v1/chat \
-H "Content-Type: application/json" \
-d '{"prompt": "Hello."}'

6️⃣ Example Response

{
  "response": "Hello! AI server received your prompt."
}

Users can now send any prompt to the AI server.

Real AI inference request.
POST /v1/chat handled by NeuroHTTP, parsed at low level and routed to a real LLM backend (LLaMA-based API).
Shows full request lifecycle, logging, and successful 200 OK response.

🔧 Important Notes

Make sure the OPENAI_API_KEY environment variable is set before starting the server.
To change the port or server options, edit include/config.h.
The server uses libcurl to communicate with the AI backend.

Benchmark Comparison

For detailed benchmark results comparing NeuroHTTP and NGINX, see benchmark.md

🧩 Visual Benchmark Evidence

Below are the live screenshots from the actual benchmark runs.

🔹 NeuroHTTP — 40,000 Connections

🔹 NGINX — 40,000 Connections

🧪 Performance Highlights

Server Conns Requests/sec Avg Latency Transfer/sec NGINX 1.29.3 10k 8,148 114ms 1.2 MB/s NeuroHTTP 10k 2,593 57ms 7.9 MB/s

💡 Insight

NeuroHTTP handles heavier, AI-rich payloads with lower latency and higher throughput per connection.

🌟 Why Star?

Low latency, high throughput
Compact C + Assembly core
Open-source & extensible

If you love high-performance AI servers, consider giving the project a ⭐ and sharing it with others.

⭐ Support the Project

If you believe in the vision of a fast, AI-native web layer, your support helps NeuroHTTP evolve. 🚀

🧬 Author

👨‍💻 GUIAR OQBA 🇩🇿
Creator of NeuroHTTP — focused on low-level performance, AI infrastructure, and modern web systems.

“Building the next generation of AI-native infrastructure — from El Kantara, Algeria.”