banner-image

Your on-prem data wants to talk.
Karaoke is listening.

Stardog Karaoke lets you have a safe and secure conversation with your on-prem data that is timely, accurate, and 100% hallucination free.

Stardog Karaoke is a turnkey Voicebox appliance that includes everything you need to bring GenAI and Knowledge Graph to your on-premise data.

Karaoke includes CPUs, GPUs, storage, and networking, solving the troublesome LLM systems integration problems that lie at the heart of GenAI.

Karaoke is fully vendor-managed by Stardog Cloud, but is resident in your data center.

High availability features ensuring uninterrupted operations, including storage and virtual clustering management.

End-to-end service and single point of accountability for both hardware and software. Comprehensive storage and backup API for seamless data management.

Choose your Karaoke model

Micro

micro

Voicebox Capacity

  • 500 concurent users
  • 250M total nodes & edges
  • 3k Voicebox tokens/sec.

System Capacity

  • 72 CPU cores
  • 0.5 TB memory
  • 12 TB NvME storage
  • 100GbE networking

GPU Capacity

  • 4k FP8 TFLOPS
  • 96 GB GPU memory

Data Services

  • Voicebox
  • Link prediction
  • External compute
  • Finetuning
  • Knowledge Graph Bootstrap from documents
  • Real-time RAG & text extraction

API

  • Voicebox API
  • Automated prompt engineering
  • LLM Gateway
  • Finetuning
  • Multi-tenant LLM
  • Vector Embeddings

Edge

edge

Voicebox Capacity

  • 1,000 concurent users
  • 1B total nodes & edges
  • 24k Voicebox tokens/sec.

System Capacity

  • 576 CPU cores
  • 3.5 TB memory
  • 60 TB NvME storage
  • 100GbE networking

GPU Capacity

  • 32k FP8 TFLOPS
  • 768 GB GPU memory

Data Services

  • Voicebox
  • Link prediction
  • External compute
  • Finetuning
  • Knowledge Graph Bootstrap from documents
  • Real-time RAG & text extraction

API

  • Voicebox API
  • Automated prompt engineering
  • LLM Gateway
  • Finetuning
  • Multi-tenant LLM
  • Vector Embeddings

Foundry

foundry

Voicebox Capacity

  • 5,000 concurent users
  • 10B total nodes & edges
  • 48k Voicebox tokens/sec.

System Capacity

  • 1152 CPU cores
  • 7 TB memory
  • 120 TB NvME storage
  • 200GbE networking

GPU Capacity

  • 64k FP8 TFLOPS
  • 1.5 TB GPU memory

Data Services

  • Voicebox
  • Link prediction
  • External compute
  • Finetuning
  • Knowledge Graph Bootstrap from documents
  • Real-time RAG & text extraction

API

  • Voicebox API
  • Automated prompt engineering
  • LLM Gateway
  • Finetuning
  • Multi-tenant LLM
  • Vector Embeddings

Enterprise

enterprise

Voicebox Capacity

  • 20,000 concurent users
  • 50B total nodes & edges
  • 96k Voicebox tokens/sec.

System Capacity

  • 2304 CPU cores
  • 14 TB memory
  • 210 TB NvME storage
  • 400GbE networking

GPU Capacity

  • 128k FP8 TFLOPS
  • 3 TB GPU memory

Data Services

  • Voicebox
  • Link prediction
  • External compute
  • Finetuning
  • Knowledge Graph Bootstrap from documents
  • Real-time RAG & text extraction

API

  • Voicebox API
  • Automated prompt engineering
  • LLM Gateway
  • Finetuning
  • Multi-tenant LLM
  • Vector Embeddings

Karaoke Integration Partners