Skip to content

nxp-appcodehub/dm-eiq-genai-flow-demonstrator

Folders and files

NameName
Last commit message
Last commit date

Latest commit

ย 

History

6 Commits
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 

Repository files navigation

NXPยฎ eIQยฎ GenAI Flow Demonstrator Package

License badge Board badge Board badge Board badge Board badge Board badge Board badge Board badge

Language badge Category badge


Overview

The NXP eIQยฎ GenAI Flow Demonstrator package showcases advanced AI capabilities running at the edge on NXP i.MX9 and i.MX8M application processors. This demonstrator includes two powerful AI applications:

  1. eIQ GenAI Flow - Conversational AI Pipeline
  2. Vision Language Model (VLM) - Visual Question Answering

Both applications are optimized for edge deployment and demonstrate real-world AI use cases on resource-constrained devices.


๐Ÿ“ฆ Package Contents

1. eIQ GenAI Flow - Conversational AI

A complete conversational AI pipeline integrating:

  • Wake-Word Detection (VIT)
  • Speech-to-Text (Whisper, Moonshine)
  • Retrieval-Augmented Generation (RAG)
  • Large Language Model (Danube 500M)
  • Text-to-Speech (VITS)

๐Ÿ“– Detailed Documentation: eiq_genai_flow/README.md

Key Features:

  • Multi-turn conversations with wake-word activation
  • Knowledge-base enhanced responses via RAG
  • Multiple input modes (voice, keyboard, GUI)
  • NPU acceleration support (i.MX95)
  • 900+ TTS voice options

2. Vision Language Model (VLM)

Visual question answering system for image understanding:

  • Multimodal AI combining vision and language
  • Image captioning and visual reasoning
  • Edge-optimized inference

๐Ÿ“– Detailed Documentation: vlm/README.md

Key Features:

  • Describe and analyze images
  • Answer questions about visual content
  • Optimized for edge deployment

๐Ÿš€ Quick Start

Prerequisites

  • Hardware: NXP i.MX95, i.MX943, i.MX93, i.MX91, i.MX8MP, i.MX8MM, or i.MX8MN board
  • OS: NXP Linux BSP (L6.12.49-2.2.0 or later recommended for i.MX95)
  • Python: Version 3.13
  • Storage: At least 16GB free space

Installation

  1. Transfer the package to your i.MX device:

    scp -r dm-eiq-genai-flow-demonstrator root@<imx-device-ip>:/root/
  2. Install eIQ GenAI Flow:

    cd eiq_genai_flow
    ./install.sh

    See eiq_genai_flow/README.md for detailed installation options.

  3. Install VLM:

    cd ../vlm
    ./install.sh

Running the Demonstrators

Conversational AI:

cd eiq_genai_flow
python3 eiq_genai_flow.py -i vasr -o tts -m danube-500M-q8

Vision Language Model:

cd vlm
./launch.sh

๐Ÿ“‹ Platform Support

Platform eIQ GenAI Flow VLM NPU Acceleration
i.MX95 โœ… Full โœ… โœ… (LLM)
i.MX943 โœ… Full โœ… โŒ
i.MX8MP โœ… Full โœ… โŒ
i.MX93 โœ… Partial* โœ… โŒ
i.MX91 โœ… Minimal** โœ… โŒ
i.MX8MM/MN โœ… Minimal** โœ… โŒ

* Lighter models recommended (danube-500M-q4, moonshine-tiny)
** RAG-only mode (no LLM generation)

See platform recommendations for detailed configuration guidance.


๐Ÿ“š Documentation


๐Ÿ”ง Configuration

Each component has its own configuration:

  • eIQ GenAI Flow: Edit eiq_genai_flow/config.py
  • VLM: See vlm/README.md for configuration options

๐Ÿ“Š Benchmarks

Performance benchmarks for various configurations are available at:


โš ๏ธ Demonstrator Limitations

This demonstrator package has the following limitations:

  • Session timeout: Applications automatically shut down after 1 hour of operation
  • Language support: English language only
  • Component delivery: Core AI components provided as optimized binary libraries
  • Model selection: Includes a curated subset of models optimized for target platforms
  • Model format: Models delivered in encrypted format

These limitations are designed to provide an optimal evaluation experience while showcasing AI capabilities on NXP platforms.


๐Ÿ†˜ Support


๐Ÿ“„ License

This software is proprietary to NXP and may only be used strictly in accordance with the applicable license terms.

See LICENSE.txt for complete terms and conditions.

Third-Party Licenses

  • See individual component licenses in the licenses/ directory

๐Ÿ“ Release Notes

Version Release Date Highlights
3.0 March 31, 2026 TTS streaming, new Audio Manager, module customization, VLM demonstrator v1.0
2.0 November 21, 2025 Neutron acceleration, customizable wake-word, 900+ TTS voices
1.1 June 20, 2025 i.MX8MP support
1.0 March 31, 2025 Initial release (i.MX95)

See component-specific README files for detailed release notes.


๐ŸŽฏ Use Cases

Conversational AI Applications

  • Smart home voice assistants
  • Automotive in-cabin assistants
  • Healthcare patient interfaces
  • Industrial HMI systems

Vision Applications

  • Visual inspection and QA
  • Assistive technologies
  • Educational tools
  • Content analysis

๐Ÿ—๏ธ Repository Structure

dm-eiq-genai-flow-demonstrator/
โ”œโ”€โ”€ eiq_genai_flow/          # Conversational AI pipeline
โ”‚   โ”œโ”€โ”€ README.md           # Detailed documentation
โ”‚   โ”œโ”€โ”€ config.py           # Configuration file
โ”‚   โ””โ”€โ”€ eiq_genai_flow.py   # Main application
โ”œโ”€โ”€ vlm/                    # Vision Language Model
โ”‚   โ””โ”€โ”€ README.md           # VLM documentation
โ”œโ”€โ”€ licenses/               # Third-party licenses
โ”œโ”€โ”€ LICENSE.txt            # Main license
โ”œโ”€โ”€ SBOM-*.spdx.json       # Software Bill of Materials
โ””โ”€โ”€ README.md              # This file

๐Ÿ”— Related Resources



NXP Semiconductors - Securing the Connected World

\

About

The eIQ GenAI Flow Demonstrator is a Conversational AI Pipeline application designed for NXP i.MX95 devices.

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors