Neural Pulse

Real-time LLM observability dashboard for OpenAI models.

Neural Pulse is a terminal dashboard for monitoring OpenAI responses in real time. Track token usage, latency, throughput (TPS), API cost, and live streaming output in a polished, htop-style interface.

Demo

Download demo video

The pipeline is an illustrative visualization inspired by transformer inference. It represents the request lifecycle and streaming activity, not the model's internal activations.

Quick Start

export OPENAI_API_KEY=sk-your-key

npx neural-pulse "Explain artificial intelligence"

Features

• Live streaming OpenAI responses
• Real-time token usage, TPS, latency, and API cost
• Timestamped event log
• Fixed dashboard with no terminal scrolling
• Interactive prompt sessions
• Compact layout for smaller terminals
• Cross-platform support (macOS, Linux, Windows Terminal, Warp, iTerm2)

Installation

Run without installing

npx neural-pulse

Install globally

npm install -g neural-pulse

Run from source

git clone https://github.com/SoorajSundar1505/neural-pulse.git
cd neural-pulse
npm install
cp .env.example .env
# Add your OPENAI_API_KEY
npm start

Authentication

Neural Pulse uses your own OpenAI API key.

Requests are sent directly to the OpenAI API. Your usage, billing, and rate limits remain on your own OpenAI account.

Configuration

Either export your API key:

export OPENAI_API_KEY=sk-your-key

or create a .env file:

OPENAI_API_KEY=sk-your-key
OPENAI_MODEL=gpt-4o-mini

| Variable | Default | Description | |----------|---------|-------------| | OPENAI_API_KEY | — | Required OpenAI API key | | OPENAI_MODEL | gpt-4o-mini | Default model | | NEURAL_PULSE_FPS | 30 | Dashboard refresh rate | | NEURAL_PULSE_ANIM_MS | 100 | Pipeline animation speed | | NEURAL_PULSE_MAX_TOKENS | (unset) | Optional completion token limit | | NEURAL_PULSE_ASCII | 0 | Force ASCII mode |

.env is automatically loaded from the current working directory.

Usage

# Interactive mode
neural-pulse

# Run with a prompt
neural-pulse "Explain transformers"

# Choose a model
neural-pulse --model gpt-4o-mini "Write a haiku about GPUs"

# Help
neural-pulse --help

# Version
neural-pulse --version

Keyboard Shortcuts

| Key | Action | |-----|--------| | P / Ctrl+P | New prompt | | Q / Ctrl+C | Exit |

Supported Models

Neural Pulse works with any OpenAI model your API key can access.

| Model | Input / 1M | Output / 1M | |-------|------------|-------------| | gpt-4o-mini | $0.15 | $0.60 | | gpt-4o | $2.50 | $10.00 | | gpt-4.1 | $2.00 | $8.00 | | gpt-4.1-mini | $0.40 | $1.60 | | gpt-4-turbo | $10.00 | $30.00 | | gpt-3.5-turbo | $0.50 | $1.50 |

Versioned model names are automatically matched (for example gpt-4o-mini-2024-07-18).

Cost Calculation

cost = (inputTokens × inputPrice + outputTokens × outputPrice) / 1,000,000

During streaming, token counts and cost are estimated in real time. When the OpenAI API returns usage information, Neural Pulse reconciles the estimates with the final reported token counts.

Requirements

• Node.js 18+
• OpenAI API key
• Interactive terminal (minimum 76 columns)

Why Neural Pulse?

Neural Pulse helps developers observe OpenAI requests in real time by tracking:

• Token usage
• Tokens per second (TPS)
• Latency
• API cost
• Streaming output
• Request lifecycle

It's designed for developers building, debugging, and demonstrating LLM applications.

Contributing

Contributions are welcome.

See CONTRIBUTING.md and the Code of Conduct.

License

MIT © Suraj S