simulate-run

name

simulate-run

description

Generate the demo workspace at {workspace_root}/simulate-run-demo/ with full /discover + /deliver + /learn artifacts following the latest plugin schema (Phase 8 PR publish, pr_urls.json, learn runs, stacks docs, architecture diagrams). Each invocation wipes and recreates the workspace in place — fixed run IDs are reused, so no timestamped run dirs accumulate. Spawns the site-view pointed at the demo by default; pass --no-ui for headless generation. Zero agent cost.

/simulate-run

Fabricates a fully-populated demo workspace AND spawns the site-view so the project drawer can be exercised end-to-end without spending agent tokens. Useful for:

• UI regression checks after editing site-view/server.js or index.html
• Demoing the plugin to teammates without running real features
• Onboarding to the pipeline-view tooling

Usage

/simulate-run [--no-ui] [--port=<n>] [--keep] [--step-ms=<n>]

Flags:

• --no-ui — generate the demo workspace files only; do NOT spawn the site-view. Use this for headless regression tests or scripted setup.
• --launch-ui — explicit form of the default. Kept for backwards compatibility — it's a no-op now since UI launch is the default.
• --port=<n> — initial port for the UI (default 5173; the server auto-increments if busy).
• --keep — skip the wipe step. Useful for incremental edits to specific files between runs.
• --step-ms=<n> — milliseconds between timeline steps for the active /deliver run (deliver_a / "Bulk Upload"). Default: 1500 when the UI is on (the default), 0 when --no-ui is set. Set --step-ms=0 to force static mode (write the completed run all at once); set --step-ms=150 for fast headless tests.

Modes

Live mode (default — UI on, --step-ms defaults to 1500): the historical artifacts (/discover, /learn runs, deliver_b) are written all at once, but the active /deliver run (deliver_a — Bulk Upload) is initialized in PENDING state. The script then steps through ~22 timeline events — phase starts/ends and parallel implementation/review tasks — appending to checkpoints.jsonl and rewriting scratchpad.md on each step. The site-view's fs.watch triggers an SSE broadcast on every change, so the browser animates forward in real time.

Static mode (default with --no-ui, or any --step-ms=0): every artifact is written all at once with backdated timestamps. If the UI is launched in this mode (e.g. /simulate-run --step-ms=0), it mounts on a fully-COMPLETED run.

Behavior

Each invocation:

1. Wipes {workspace_root}/simulate-run-demo/ (unless --keep).
2. Generates the workspace skeleton: config.json, context/platform.md, context/audit-findings.md, context/diagrams/architecture.mmd, context/diagrams/architecture-overview.mmd, history/learn-log.md.
3. Generates 1 fixed /discover run with checkpoints, scratchpad, and report.
4. Generates 2 fixed /deliver runs (one with Phase 8 PRs + pr_urls.json, one without).
5. Generates 2 fixed /learn runs (PR-based + run-based source modes).
6. If a real workspace exists with a recent /deliver run, copies that run's scratchpad.md / report.md / pr_urls.json / checkpoints.jsonl into the demo as a third run (with run_id renamed) — so the demo reflects realistic feature output when available, fully synthesized when not.
7. Fixed run IDs (e.g. 2026-04-25-110000-bulk-upload) are reused on every invocation, so the demo workspace never accumulates timestamped junk.

Instructions

When the user invokes /simulate-run:

1. Pass through every CLI flag the user provided to the Node script.
2. Launch via Bash:
   • Default (UI on): run_in_background: true so the spawned site-view keeps serving. The script writes the historical runs immediately, then steps the live deliver_a run for ~33 seconds (22 events × 1500ms) so the browser animates.
   • With --no-ui: foreground — the script generates files and exits.
3. Report the demo workspace path, the URL the site-view is serving on, and which artifacts were produced.

node {plugin_dir}/scripts/simulate-run.js [--no-ui] [--port=<n>] [--keep] [--step-ms=<n>]

What gets exercised in site-view

After running (default — UI on), opening the project drawer should show every section populated:

• Workspace info + repos (4 demo repos)
• Audit summary with critical/high/medium/low pills
• Last discover run + per-phase token breakdown
• Delivered features (2 or 3) with PR links on the published one
• Learnings captured (2) with finding counts
• Architecture diagrams (Mermaid)
• Platform context (rendered markdown)

The site-view's lazy loaders are also exercised: the deliver-run detail and learn-run detail endpoints are hit when the user expands a row.

Cleanup

The demo workspace persists between runs. To remove it:

rm -rf {workspace_root}/simulate-run-demo/

Or rerun /simulate-run and just don't use it.