Makerchip VS Code Extension — Monorepo

This is an npm workspaces monorepo containing three published/publishable packages:

Package	npm / VS Code id	Purpose
packages/tlv-grammar	@rweda/tlv-grammar	TL-Verilog TextMate grammar and language configuration. Framework-neutral — reused by VS Code, Monaco, Shiki, and GitHub Linguist. No runtime code.
packages/tlv-extension	redwood-eda.tl-verilog	TL-Verilog language extension: .tlv/.TLV registration, grammar, snippets, semantic tokens, hovers, and (desktop-only) module instantiation. Grammar files are copied from tlv-grammar at build time — edit only in tlv-grammar/.
packages/makerchip-extension	redwood-eda.makerchip	Makerchip IDE integration: server-connected webview (compile/simulate, diagram, waveform, Nav-TLV, VIZ), GitHub Copilot LM tools, @makerchip chat participant, local compile cache.

Supported use models

Three use models drive architectural decisions:

1. Desktop Makerchip — full IDE + Copilot + local compile cache + local tools (webview, module instantiation, experimental SandPiper SaaS / Verilator / GTKWave buttons).
2. Web Makerchip (vscode.dev / github.dev) — server-hosted IDE + web-safe language features; no local FS or tool spawning.
3. Editor-only TL-Verilog — highlighting, semantic tokens, snippets, and hovers with no Makerchip server or network, fully offline, both hosts. Served by the standalone tl-verilog extension. The grammar is also consumed outside VS Code entirely (GitHub Linguist, Shiki, Monaco).

Package coupling

makerchip declares a hard extensionDependencies on redwood-eda.tl-verilog. VS Code auto-installs tl-verilog alongside Makerchip and won't activate Makerchip without it. makerchip contributes no language definition or grammar itself — it only references the tlverilog language id in when clauses.

The language layer is kept as a separate, independently installable extension so that editor-only TL-Verilog (use model 3) remains viable without pulling in the Makerchip IDE.

Development

Prerequisites

• Node.js ≥ 18, npm (workspaces used for shared deps)
• Install all dependencies: npm install from the repo root

Building

# One-shot build of all packages
npm run compile

# Watch mode for all packages (started automatically on folder open in VS Code)
# Both makerchip-extension and tlv-extension rebuild on save.
# See .vscode/tasks.json

Launching the Extension Development Host

Always launch the VS Code Extension Development Host via the ./launch script. A convenience symlink at the repo root (./launch) forwards to the real script in packages/makerchip-extension. Its single argument selects the sandhost you develop against — a deployed URL or a local mono clone:

./launch                             # Deployed default server (beta.makerchip.com)
./launch https://makerchip.com/v123  # Deployed server at an explicit URL
./launch example_videos              # A local mono clone (mono_example_videos)

For a clone it wires up everything (SandHost, Cloudflare tunnel, dedicated workspace, dedicated VS Code profile) and loads both makerchip-extension and tlv-extension from source, satisfying the hard extensionDependencies without a Marketplace publish. Clone resolution, tunnel ownership/teardown, the bin/reap reaper, and the full per-clone workflow are documented in the package README.

Co-developing both extensions

The "Run Extension" launch config (F5) (default deployed sandhost only) and the ./launch script both pass two --extensionDevelopmentPath args so edits to either extension are picked up together. The default build task (watch: all) runs tsc -watch for both in parallel.

Important: the Extension Development Host does not hot-reload. After tsc finishes recompiling (watch terminal shows "Found 0 errors. Watching…"), reload the dev-host window to activate the new code.