Getting Started

Developer Tools

Eta ships a full developer-tools ecosystem: a VS Code extension with LSP, debugger, and heap inspector; a Jupyter kernel for interactive notebooks with rich output; and a built-in sampling profiler accessible from the CLI, REPL, and notebooks.

VS Code Extension

The Eta VS Code extension provides first-class editing support powered by the native eta_lsp language server and eta_dap debug adapter.

LSP — real-time diagnostics, hover docs, go-to-definition, completion, rename, and signature help.
Debugger (DAP) — breakpoints, stepping, call-stack inspection, and variable watch.
Heap Inspector — live view of the GC heap with per-object detail.
Disassembly View — per-function bytecode listing with current-PC highlighting during debug pauses.
Test Runner — integrated eta_test runner with pass/fail markers in the editor gutter.

VS Code Extension reference

Jupyter Notebooks

eta_jupyter is a native Jupyter kernel for Eta. Install it once and use it from JupyterLab or classic Notebook just like any other kernel.

Rich output — tensors, fact-tables, DAGs, actor state, and Vega-Lite plots render as interactive widgets.
Cell magics — %time, %bytecode, %load, %%prof, and more.
Inline profiling — %%prof profiles a single cell and renders a flamegraph inside the notebook output.
Comm widgets — live heap inspector, disassembly view, and actor-state panel without leaving the notebook.

Eta notebook — basic cell execution and output — Basic Eta notebook execution and cell output.

Eta notebook — rich rendering output — Rich rendering: tensors, tables, and plots rendered inline.

Jupyter kernel reference

Profiler

Eta has a built-in runtime profiler with two modes:

Sample mode (default) — low-overhead sampled call stacks, outputs speedscope JSON for flamegraph visualisation.
Trace mode — exact call/return accounting, higher overhead, suitable for precise attribution.

The profiler is accessible from three surfaces:

CLI — eta prof run, eta prof report, eta prof merge, eta prof view.
REPL — the :prof meta-command profiles the next submission.
Notebook — the %%prof cell magic profiles one cell and renders an inline flamegraph.

# Run sampled profiling and open in speedscope
eta prof run --out profile.speedscope.json path/to/program.eta
eta prof view profile.speedscope.json

# Trace mode with pretty-print report
eta prof run --mode trace --format pretty path/to/program.eta

Profiler reference