> ## Documentation Index
> Fetch the complete documentation index at: https://dragonwingdocs.qualcomm.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Tools and acceleration

> Review Ubuntu tools, AI runtimes, multimedia paths, and developer utilities for Dragonwing platforms.

Ubuntu on Qualcomm® IoT platforms is more than an operating-system image. The documentation also points to tools and acceleration workflows that help you validate hardware, run models, build applications, and profile performance on target devices.

## Tooling areas

<CardGroup cols={2}>
  <Card title="Remote development" icon="terminal">
    Use SSH and remote editor workflows to edit, build, run, and debug directly on the target.
  </Card>

  <Card title="Containers" icon="box-open">
    Use container workflows where they are available for AI, development, or runtime isolation.
  </Card>

  <Card title="Profiling" icon="chart-line">
    Profile models and workloads to understand CPU, memory, accelerator, and runtime behavior.
  </Card>

  <Card title="Performance utilities" icon="gauge-high">
    Use Qualcomm performance utilities where documented for the target and release.
  </Card>
</CardGroup>

## Acceleration paths

| Area               | Documentation starting point                                                                                                  |
| ------------------ | ----------------------------------------------------------------------------------------------------------------------------- |
| AI Hub models      | [AI Hub workflow](ai-workflows/ai-hub)                                                                                        |
| LiteRT             | [LiteRT workflow](ai-workflows/lite-rt)                                                                                       |
| ONNX Runtime       | [ONNX Runtime workflow](ai-workflows/onnxruntime)                                                                             |
| PyTorch on NPU     | [Run PyTorch models on NPU](ai-workflows/run-py-torch-models-on-npu)                                                          |
| LLM/VLM containers | [LLM/VLM setup](ai-workflows/llm-vlm-qcom-setup)                                                                              |
| IM SDK             | [IM SDK workflow](ai-workflows/im-sdk)                                                                                        |
| QIR SDK            | [QIR SDK workflow](ai-workflows/qir-sdk)                                                                                      |
| Speech workflows   | [Whisper](ai-workflows/whisper), [ASR container](ai-workflows/asr-container), and [TTS container](ai-workflows/tts-container) |

## Practical guidance

* Start with a known-good device setup before running accelerated workloads.
* Install the required packages listed for your board and workflow.
* Validate camera, audio, display, network, and storage dependencies before debugging an AI or multimedia application.
* Keep a record of image version, package versions, model files, and command lines when comparing results.

<Note>
  Some tools and acceleration workflows are device- or release-specific. If a linked page is not present in your device navigation, use the closest supported EVK workflow or the release notes for availability guidance.
</Note>