Tera Language

Language support for Tera (.tera) — a tensor algebra & neural network DSL.

Syntax highlighting, diagnostics, and IntelliSense powered by a built-in language server.

Features

• Syntax highlighting for tensors, models, functions, and control flow
• Diagnostics — type errors and undefined names reported as you type
• IntelliSense — completions for builtins, tensor methods, and modules
• Snippets for common patterns (models, layers, training loops)
• Bracket matching, auto-indent, and comment toggling

Example

x = tensor([[1, 2], [3, 4]])
w = randn([2, 3])
y = (x @ w).relu()

model MLP(input: int, hidden: int, output: int):
  fc1 = Linear(input, hidden)
  fc2 = Linear(hidden, output)

  forward (x: Tensor) -> Tensor:
    x = fc1(x).relu()
    return fc2(x)

net = MLP(2, 4, 1)
print(net(randn([8, 2])).shape)

Requirements

VS Code ^1.94.0. The extension activates automatically when you open a .tera file.

Settings

License

MIT