Palette Neat

Palette Neat is the SiMa.ai software development toolkit for building AI applications on Modalix. It includes the development environment, runtime library, model tooling, and DevKit validation workflow. Together, these components support the full development path: prepare a model, build an application, and validate the result on Modalix hardware.

Use this overview to understand the main parts of Palette Neat and choose the right setup, model preparation, or application development path.

Software stack

Palette Neat includes four software components and the Modalix DevKit:

• Neat Development Environment — the containerized environment for cross-compilation, DevKit pairing, agent-assisted workflows, and local validation.
• Neat Library — the C++ and Python library and runtime for loading models, executing pipelines, and coordinating work across Modalix compute resources.
• Model Compiler — the optional component for model quantization, compilation, and validation.
• LLiMa — the GenAI toolkit, included with the development and model tooling, for compiling, testing, and benchmarking GenAI models.
• Modalix DevKit — the target hardware where compiled model artifacts and Neat Library applications run and are validated.

Developer journey

If you are new to Palette Neat, work through these sections in order: confirm compatibility, set up the Neat Development Environment, install or update the Neat Library, prepare a model, and then build the application that uses it.

Start Here

Prepare your host machine, Neat Development Environment, and DevKit for local development and hardware validation.

• Check CompatibilityConfirm supported host, Neat Development Environment, DevKit software, Neat Library, and Model Compiler versions.
• Neat Development EnvironmentInstall the SDK, pair a DevKit, and run on hardware with dk.
• Neat LibraryInstall or update the runtime and PyNeat on a DevKit or paired environment.

Model Preparation

Turn trained models into deployable artifacts that run on Modalix hardware.

• Compile a ModelCompile pretrained ONNX vision models or GenAI models for Modalix.
• Use a Precompiled ModelStart quickly with a ready-to-run model artifact.
• GenAI with LLiMaCompile, test, and benchmark GenAI models on Modalix.

Build an App

Use the Neat Library to run models and compose production application pipelines.

• Hello Neat!Run your first Neat application and verify the development workflow.
• Develop AppsBuild AI applications with the Neat Library using C++ or PyNeat.
• TutorialsFollow guided examples for real Neat application patterns.

Tools & Reference

Use supporting tools and reference material when you need more detail.

• Toolssima-cli, Model Zoo and Insight.
• ReferenceBrowse APIs, troubleshooting, environment variables, data formats, and release notes.
• TroubleshootingFind fixes for setup, runtime, and pipeline issues.