NIM Kit

A toolkit to explore NVIDIA NIMs locally — from LLMs to 3D and beyond.

Quick Start

1. Start the backend services:

   docker compose up

2. Start the frontend development server:

   cd frontend
   yarn && yarn dev

3. Configure NVIDIA API (optional): Copy .env.sample to .env and add your NVIDIA_API_KEY for enhanced functionality.

Supported NIM Services

Service Type	Model	Description
LLM	Meta Llama 3.1 8B Instruct	Advanced language understanding and text generation
LLM	DeepSeek R1 Distill Llama 8B	Enhanced reasoning and coding capabilities
Image Generation	FLUX.1 Schnell	Fast, high-quality image generation
Image Generation	FLUX.1 Dev	State-of-the-art image generation suite
Image Generation	FLUX.1 Kontext Dev	Multimodal in-context image generation and editing
3D Generation	Microsoft Trellis	Generate 3D assets from text or images
Speech Recognition	NVIDIA Parakeet CTC 0.6B ASR	English speech-to-text transcription
Speech Enhancement	NVIDIA Studio Voice	Enhance speech quality and correct audio degradations
OCR	Baidu PaddleOCR	Optical character recognition and table extraction

Architecture

• Backend: FastAPI with Celery for async task processing
• Frontend: Nuxt.js with Vue 3, Tailwind CSS and shadcn-ui
• Database: Redis for caching and task queuing
• Monitoring: Flower for Celery task monitoring

NIM Observability

Monitor your NIM deployments with built-in observability tools including Prometheus metrics, Grafana dashboards, and alerting rules. Navigate to the nim-observability/ directory to set up comprehensive monitoring for your NIM services.