elizaOS
LifeOps Personal Assistant MVP — coordination #14407
Replies: 10 comments
-
New shared QA tooling:
|
Beta Was this translation helpful? Give feedback.
-
|
Human-only work dossiers filed (personal-corpus scenario program, epic #14747):
Board hygiene card: #14750. All dossier issues are on Project 15 (Status=Todo). |
Beta Was this translation helpful? Give feedback.
-
Session coordination summary — headless MVP-UI/testing pass (2026-07-05)Posting to the central coordination point per the workstream process. This session ran an opus-implement → fable-review → fix → re-verify loop across the code-tractable board-15 gaps, prioritising test/device architecture first, then user-facing UI. Shipped (18 PRs; 0 duplicate PRs against the concurrent swarm; 5 real defects caught+fixed by the review loop)
Defects the fable loop caught (then fixed)A nonexistent Docker image tag ( Visual QA (real pixels)19 rendered states captured + analyzed ( Design-doc verification (MVP.md)Cross-referenced MVP.md §"what's weak" against develop source: all five concrete correctness bugs it names are fixed (weather hardcode, ungated 5s poll, English-only STT, Android stale-install, GenUI-first prompt) — details on #14356. The doc's thesis holds; the residual is verification, not features. Honest status / coordination askThe remaining open board items are live-model + device verification — persona packs A1/A2/B2 live trajectories, the live onboarding-journey scenario, child-voice/student-deadline scenarios, a green scheduled |
Beta Was this translation helpful? Give feedback.
-
Per-issue review — all 23 open MVP-block issues (individual verdict)Reviewed each open board-15 UI/testing item individually and classified by its actual remaining blocker. Result: 0 are "code-tractable but unaddressed" — every open item is code-landed-awaiting-human-verify, genuinely resource-gated, or an owner decision. A. Code landed; "open" only pending human sign-off or a linked live leg (agent-done)
B. Genuinely
|
| # | Blocked on |
|---|---|
| #14348 home resting-state e2e | home rendered with populated keeper widgets (real backend) |
| #14351 / #14352 / #14353 / #14354 / #14355 | live-model persona/onboarding/goal scenario trajectories, hand-read |
| #14356 fresh-user view set + F1 | live-model — MVP.md design-doc findings posted to the issue |
| #14357 resurrect app-live-e2e | real cloud login + real LLM (CI creds) |
| #14358 device e2e cloud sign-in | physical device + cloud |
| #14370 / #14371 voice benchmark + e2e | Railway services + real audio/device |
| #14414 mobile build/cold-boot perf | device profiling |
C. Owner product decision
- [onboarding] P2 — Remove the one-time character-select landing after cloud-only onboarding (owner confirm, then delete) #14362 remove the post-onboarding character-select landing — fix(#13371): make character select post-onboarding only #13396 made it deliberate, so it needs owner confirm before deleting.
One concrete cleanup: #14373 is satisfied on develop by the merged #14511 but the issue stayed open (PR used fix(#…) not Closes) — safe to close.
Everything code-tractable is landed and evidenced (19 session PRs incl. the device-found iOS blue-tint fix #14826 + the audit:live harness #14818). The rest is the board's own needs-human queue + the human-verify "Done" gate.
Beta Was this translation helpful? Give feedback.
-
|
MVP evidence-state refresh — 2026-07-09 I re-ran the current audit loop from latest available node packages/scripts/audit-mvp-evidence-matrix.mjs
node packages/scripts/check-mvp-board-readiness.mjs --json --no-fail
node packages/scripts/check-lifeops-persona-catalog-coverage.mjs --unverifiedCurrent state:
The remaining MVP closeout work is evidence, not another unowned implementation lane. The missing proof is concentrated in:
I also checked this shell for the live credentials needed to produce the trajectory evidence: So I cannot honestly attach the required live-LLM or cloud evidence from this environment. The correct next move is a credentialed evidence pass, not closing these issues from static code state. Once credentials/human artifacts are available, the audit commands above should be rerun first, then each issue should get its own inline proof bundle before labels/status are cleared. |
Beta Was this translation helpful? Give feedback.
-
Project 15 audit correction — #15748 / PR #15750The 2026-07-09 closeout snapshot is false-green because the scripts scope by the optional Live independent reconciliation now shows 36 open Project 15 issues, not 23. Readiness and evidence-matrix output match the independent The repaired readiness gate currently fails for:
Board audit totals: 195 project issues, 162 with |
Beta Was this translation helpful? Give feedback.
-
|
Atomic Project 15 closeout PR #15763 is open. One live snapshot now feeds board/readiness/evidence analysis and validates exact project/open/closed coverage plus 37/37 readiness/evidence parity. Current result is integrity PASS but MVP NOT READY: 33 human-gated, 4 actionable (#13406, #13631, #15753, #15758), 4 closed-not-Done, and status mismatches on #14793/#14797/#14864. This replaces the rate-limit-prone three-command snapshot documented earlier. |
Beta Was this translation helpful? Give feedback.
-
|
MVP QA lane update (2026-07-09): the original active code PRs #15754, #15763, #15768, and #15774 are merged. A fresh current-renderer app audit completed 241/241 Playwright tests with 240 DOM rows = 240 PNGs = 240 OCR rows, 0 broken, 0 needs-work, and 0 OCR regressions. Manual review confirmed current Polymarket/wallet across four layouts; #15780/#15781 were closed as stale evidence. Root provenance fix #15805 is merged. Review found one retry-lifecycle residual—worker-level beforeAll cleanup could erase earlier passed screenshots after a Playwright retry—fixed in #15863 (CI running) with real runner integration and unsafe-root guards. Plugin bundle provenance hardening is in #15854 (CI/evidence rerun), covering stale stub declarations, real-bundle headers, missing-bundle failure, and audit wiring. Evidence: https://github.com/elizaOS/eliza/releases/download/pr-evidence/15857-verification.log and https://github.com/elizaOS/eliza/releases/download/pr-evidence/15857-current-polymarket-wallet.jpg. |
Beta Was this translation helpful? Give feedback.
-
|
Project 15 closeout reconciliation (2026-07-09 ET):
No cards were moved directly to Done by this lane. |
Beta Was this translation helpful? Give feedback.
-
|
MVP board audit correction (2026-07-10): I re-read all 34 open Project 15 cards that were marked Needs human review against current develop and their linked PR/evidence history. The bucket contained 18 cards with remaining agent implementation, 7 code-complete cards missing required real evidence, 8 genuinely human-owned gates, and 1 Project 12 tracker misplaced as an MVP work card. I moved the 18 engineering cards back to Ready, removed stale needs-human/needs-shaw labels, and added a card-specific remaining-work comment to each issue. I claimed #14772 and opened #15952. The current corpus verify stage was a no-op that could mark raw rows verified; #15952 disables that path and adds a fail-closed, exact-byte, hash-bound corpus verifier. It deliberately cannot claim the real corpus green until #14769 deletion/attachment approval, production canary and placeholder-registry artifacts, multimodal scope, and owner sample review exist. Audit hardening PR #15874 remains exact-head CI-bound after repeated develop rebases; its current aesthetic audit is still running. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
This is the coordination hub for the LifeOps Personal Assistant MVP. If you are picking up work on chat, onboarding, the views, or LifeOps (scheduling, calendar, reminders, goals, todos, tasks, coordination), start here.
What the MVP is
One assistant, chat-first, that runs a person's real life: scheduling, calendar, coordination, reminders, goals, todos, tasks. It serves everyone on the same interface — children, adults with ADHD/ADD/Asperger's/autism, neurotypical people, business users, and elderly people — with no therapy language and no special rails. Personas are data (owner facts, anchors, reminder intensity, quiet hours), not code branches. We prove it works by covering real life: brushing teeth twice a day, workouts, getting the report done on time for class, life goals, night-owl / irregular-sleep schedules, bills, appointments, elderly check-ins.
Guiding constraint: minimize new scope. Mostly turn what we already have into an MVP by fixing, testing, verifying, and validating the important stuff. Prefer deleting/simplifying over adding.
How we coordinate
packages/docs/ongoing-development/— the MVP doc, one research/design doc per workstream, and a status snapshot.The flow: discussion → design doc in
ongoing-development/→ issues on board 15 → PR with evidence posted inline → design/status doc updated on merge.How to claim work
packages/docs/ongoing-development/research/first.developwith evidence inline (see below). Update the design/status doc when it merges.Evidence rules (this changed)
Evidence is posted inline in the issue/PR itself — drag-and-drop, not committed to the repo. The old
.github/issue-evidence/directory is retired..mov/.webm).<details>block when long.PR_EVIDENCE.md.Workstreams
The nine workstreams, each with a design doc and a kickoff thread:
Kickoff threads
Filed issues: 55 across the workstreams — all on board 15.
Beta Was this translation helpful? Give feedback.
All reactions