10 lines
1.4 KiB
Markdown
10 lines
1.4 KiB
Markdown
---
|
|
engine: devin
|
|
cwd: /Users/sd9235/code/mygh/learning_ai_local_memory_gpt
|
|
yolo: true
|
|
lock: localmemgpt-web
|
|
timeout: 4h
|
|
---
|
|
|
|
Author the baseline audit into and then execute end-to-end the roadmap at web/docs/roadmaps/UX_INTEGRATION_LOCALMEMGPT.md, STRICTLY scoped to /Users/sd9235/code/mygh/learning_ai_local_memory_gpt/web — follow waves UX-1→UX-7 + CC.1→CC.6 in order and obey its Ground rules exactly (scope web/; --lmg-* tokens with a --bl-*→--lmg-* bridge, no new color literals; @bytelyst/* deps "*"; ESM "type":"module"; minimal lockfile; tests sacred; commit→push→tick PER ITEM with SHA + test counts; deferrals [~]+reason+table; no stray files). This app has NO adapter: UX-1 must BOOTSTRAP — create src/components/ui/Primitives.tsx, add the --bl-*→--lmg-* bridge, add @axe-core/playwright so e2e/accessibility.spec.ts runs, add the ratchet + export test. Pin dev port 3070 for e2e. PRESERVE the chat markdown/syntax-highlighting UX — UX-5 only adopts @bytelyst/rich-text if it keeps parity, else defer; defer UX-7 (no auth/feed/charts). Verify each wave: pnpm --filter @localmemgpt/web typecheck && lint && test && build && test:e2e (port 3070, no new failures vs the UX-1 baseline / no new a11y violations). Capture the e2e baseline on UX-1. Do NOT stop until every non-deferred wave + CC is committed, pushed, and ticked. End with SHAs, final test counts, and deferrals.
|