A VS Code extension that sets up the environment for TI Edge AI (TinyML) workflows. It detects, installs and updates three components, plus the CCStudio extension as step 0 in VS Code:
| Step |
Component |
What it is |
| 0 |
CCStudio extension |
The ti-development-tools.ccstudio extension, installed from the VS Code Marketplace. Shown only when the host is VS Code. |
| 1 |
Python |
A private Python 3.10 interpreter (or your own 3.10.x interpreter or virtual environment). |
| 2 |
TinyML ModelZoo |
TI's training, quantization and compilation toolchain, installed as four pip packages (tinyml-modelmaker, tinyml-modelzoo, tinyml-tinyverse, tinyml-torchmodelopt) into the Python from step 1. |
| 3 |
TinyML AI Skills |
A set of skills that teach your AI agent the TinyML workflow. |
Once all three are ready, the setup panel shows starter prompts to try in your AI agent.
Usage
- On startup the extension checks all components. A status bar item (bottom right) shows a check mark when everything is ready, or a warning with the number of components that need attention.
- Click the status bar item, or run CCStudio Edge AI Toolkit: Open Setup from the Command Palette, to open the setup panel.
- The panel walks through the steps in order. Use Install, Update or Retry on the current step, Select Python… to point at your own interpreter, and Re-check to re-run detection.
- On the first run with problems, the panel opens automatically. After that you get a dismissible notification, with a "Don't remind me" option (undo it with CCStudio Edge AI Toolkit: Re-enable Setup Reminders). When a newer package version is available on the server, you are notified once per version.
Where the skills are installed
Installing TinyML AI Skills downloads the package and unpacks it to C:\ti on Windows (~/ti on macOS and Linux), the CCStudio products folder, regardless of edgeai.installDir. The extension does not copy the skills anywhere else. Making them available to your AI agent is left to CCStudio, which copies skills from the package to each supported agent.
Agent configuration
When Python, ModelZoo and Skills are all ready, the extension writes ~/.tinyml-agent-skills/.env, which the skills read at session start. It records the Python path and the install locations of the ModelZoo packages.
- At startup it only repairs a missing or broken file. Hand edits to a working file are left alone.
- The setup panel (opening it, Re-check, and every install) overwrites the keys the extension manages so that the panel's selections win. Keys it doesn't manage are preserved.
Settings
| Setting |
Scope |
Description |
edgeai.pythonPath |
machine-overridable |
Path to the Python executable. If empty, the Edge AI Python package installed by the extension is used. The system PATH is not searched. |
edgeai.installDir |
machine |
Where downloaded packages are installed. Default: %LOCALAPPDATA%\Texas Instruments\edgeai on Windows, ~/Library/Application Support/Texas Instruments/edgeai on macOS, ~/.local/share/Texas Instruments/edgeai on Linux. |
These settings control what the extension executes and installs, so they are restricted:
machine settings can only be set in your user (or remote) settings. A repository's .vscode/settings.json cannot change them.edgeai.pythonPath can also be set per workspace (for a project-specific interpreter), but in an untrusted workspace only your user-level value is used.
Proxy
Downloads (package server and pip) use the proxy from your user VS Code http.proxy setting if set, otherwise the HTTPS_PROXY / HTTP_PROXY environment variables. A workspace's http.proxy is ignored. Credentials in a proxy URL are masked in the setup log.
Install sources
All packages are downloaded from https://dev.ti.com/tirex.
Version compatibility
compatibility.json at the repository root defines the package IDs, the minimum (and for Python, maximum) supported versions, the pip package names and the package server. Edit it to change what the extension requires.
