docs: fix drift in Local Setup

[jack-arturo]

Automated docs-accuracy audit of development/local-setup. This page was 61 days since its last review (the longest gap on the docs site), and the drift is concentrated in fabricated env vars and stale build scripts.

Claim	Current state	Fix	Evidence
FLASK_ENV (default production, "development enables debug") and LOG_LEVEL (default INFO) are documented config vars, with a "Enable development mode for verbose logging and auto-reload" snippet claiming FLASK_ENV=development gives stack-trace error pages and auto-reload	Neither variable is read anywhere in the repo — a repo-wide search for FLASK_ENV, FLASK_DEBUG, and LOG_LEVEL across .py/.yml/.toml/.sh/Makefile returns nothing. app.run() hardcodes debug=False, so there is no debug mode and no auto-reload to enable.	Dropped both table rows; replaced the snippet with a caution stating the flags are inert and debug=False is hardcoded	runtime_wiring.py#L109
npm prebuild is node scripts/sync-template-versions.mjs; postbuild is node scripts/build-openclaw-plugin-package.mjs && chmod +x dist/index.js	prebuild now also regenerates the shared memory policy first (tsx scripts/sync-memory-policy.ts && …); postbuild has no chmod step	Both rows corrected	package.json scripts.prebuild / scripts.postbuild
Node.js prerequisite: "Version 20.0.0 or higher"	engines.node is ^20.19.0 || ^22.13.0 || >=24 — 20.0.0–20.18.x, 21.x, and 23.x all fail the check	Prerequisite now quotes the real range and calls out the excluded versions	package.json engines
Startup sequence diagram: "Schedule decay (1h) / creative (1h) / cluster (6h) / forget (1d)"	Defaults are 86400 (1d), 604800 (7d), 2592000 (30d), and 0 — forgetting is disabled by default, not daily	Diagram now reads decay (1d) / creative (7d) / cluster (30d) / forget (disabled)	config.py#L31-L42
Dev-environment diagram names the optional API container memory-service	docker-compose.yml defines falkordb, qdrant, and flask-api — there is no memory-service service	Diagram node renamed to flask-api	docker-compose.yml#L37

Verified against: verygoodplugins/automem@8ff266e
Verified against: verygoodplugins/mcp-automem@2816beb

Questions

None — every fix above is a direct source contradiction.

Unverified

• The scripts/bootstrap_dev.sh tip is pinned to ed36b98e, an old SHA. I checked git log ed36b98e..HEAD -- scripts/bootstrap_dev.sh and the file has not changed since, so the permalink still resolves to identical content. Left as-is rather than churning the link.
• Checked and correct as written: Python 3.12 via .python-version (3.12.13), the pinned dev-tool versions (pytest==8.3.4, black==24.8.0, isort==5.13.2, flake8==7.1.1, pre-commit==4.0.1), make dev → docker compose up --build, make fmt → black + isort, the unit pytest marker, the remaining config-table defaults, and every other npm script row.

Follow-ups

• The prod-dependency bullet lists nltk and rich as installed via -r requirements.txt, but both live in requirements-dev.txt only (nltk==3.9.1, rich>=13.0.0). requests, python-dotenv, and spaCy in that same sentence are correct.
• Two places say Flask "binds to 0.0.0.0:8001", but app.run(host="::", …) binds the IPv6 any-address (dual-stack) rather than 0.0.0.0. Cosmetically minor, but the literal string is wrong.

Both were beyond the 5-fix cap.

---

Generated by Claude Code

[cloudflare-workers-and-pages]

Deploying automem-website with    Cloudflare Pages

Latest commit:	bb4204a
Status:	✅  Deploy successful!
Preview URL:	https://8b3894fe.automem-website.pages.dev
Branch Preview URL:	https://docs-audit-development-local-16v0.automem-website.pages.dev

View logs