Skip to content
| Marketplace
Sign in
Visual Studio Code>Debuggers>Shape TellerNew to Visual Studio Code? Get it now.
Shape Teller

Shape Teller

Ricky Zhu

|
859 installs
| (0) | Free
Display shapes of tensors/arrays when over over a python variable by asking LLMs for help.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

ShapeTeller

An LLM-powered vscode extension that tells the shape of arrays and tensors when hovering over a python variable.

[!CAUTION] Since we make use of LLMs, the resulting shapes may not be accurate due to the inherent problem of hallucination. If you feel doubtful about the result, or want to be 100% accurate, you'd better run the code and print the shapes.

[!NOTE] This extension is considered to be completed. I will continue publishing new versions just to show that I keep an eye on it. If you have any proposals/problems, please open issues or pull requests, I will check them.

Installation

Install from VScode Extension Marketplace

How to use

Step 1: choose your favorite platform and model (shapeteller.platform.modelId) and set the API key (shapeteller.platform.apiKey).

Step 2: Turn on the extension when you want to use from the command palette by typing shapeteller: turn on.

Step 3: Hover over a tensor/array variable to see its shape.

Step 4: Turn off the extension when you don't want to use it from the command palette by typing shapeteller: turn off, avoiding unnecessary API calls.

Use your own LLM

If you deployed your own model using Ollama or vllm, which support OpenAI-compatible API, then you can use it as well. Just set the shapeteller.customised.modelId and shapeteller.customised.apiEndpoint in the settings. For example, if you run Ollama to serve a phi3:mini model, and your API endpoint is (by default) http://localhost:11434/v1:

[!NOTE] If you set your shapeteller.customised.modelId and shapeteller.customised.apiEndpoint, then the shapeteller.platform.modelId and shapeteller.platform.apiKey would be ignored. That says, the extension will prioritise your customised model over the commercial ones.

General Settings

There are some more general settings, see the screenshot above.

More information

  • design note

Credits

  • Idea directly brought by @msaroufim's PyTorch extension.
  • Extension icon created by Nadiinko - Flaticon
  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2025 Microsoft