feat(agent): add llm assistant chat with api tools (task 88)

OpenAI-compatible client with mock fallback, tool-calling loop, and /assistant UI.

Tools map to /api/v1/ endpoints per ADR 0006 direct-tools decision.

STATUS.md

@@ -19,6 +19,8 @@ Living progress tracker. Update at the end of each task. Codex and Claude Code b
 
 - **86 — Journal module** (ADR 0005 accepted `8e2ddd6`). Per-user `journal_entries` schema + migration `0020_journal_entries.sql`; mood catalog (multi-select), optional stress/pills, rich-text body; index with calendar dots, entry editor, Recharts mood tracker, insights (streak/trends/correlations); `/api/v1/journal/entries` CRUD with bearer token auth; OpenAPI updated. Unit tests: `journal-analytics.test.ts`. E2E: `tests/e2e/journal.spec.ts`. Run `pnpm db:migrate` for migration `0020`.
 
+- **88 — LLM agent chat** (ADR 0006). `/assistant` chat UI; OpenAI-compatible LLM client (`LLM_BASE_URL` / `LLM_MODEL`); mock provider when unset; direct tool schemas → `/api/v1/` HTTP calls (lists, calendar, notes, journal); `POST /api/agent/chat`. Unit tests: `agent-chat.test.ts`. E2E: `tests/e2e/assistant.spec.ts`.
+
 - **01 — Repo init & tooling** (commit `b89690a`). pnpm 10 + TS strict + ESLint flat + Prettier. All acceptance criteria green.
 - **02 — Next.js app skeleton**. Next.js 15 + React 19 + Tailwind v4 + shadcn/ui (button, card, input, dialog). `pnpm dev` serves placeholder, `pnpm build` produces `.next/standalone/`, `pnpm lint` clean. Added `.npmrc` with `node-linker=hoisted` for Windows symlink compatibility.
 - **03 — Drizzle + Postgres setup**. drizzle-orm + postgres driver + drizzle-kit wired up. `src/modules/_core/schema.ts` declares `users`, `households`, `household_members`. `docker-compose.dev.yaml` starts Postgres 16. `drizzle/0000_silent_magma.sql` generated and applied. `tsc --noEmit` passes.
@@ -65,11 +67,11 @@ Phase 9 — Post-v0.1 (see `docs/superpowers/specs/2026-07-03-backlog-triage-des
 2. ~~API foundation: task 87 (+ ADR 0006)~~ — done
 3. ~~Shared rich-text + notes overhaul: task 85 (+ ADR 0004)~~ — done
 4. ~~Journal: task 86 (+ ADR 0005), including journal API endpoints~~ — done
-5. LLM agent chat: task 88
+5. ~~LLM agent chat: task 88~~ — done
 
 P2/P3 backlog is filed on Gitea only (no task briefs yet) — see `docs/issues-map.md` designs 7–9, 11–12, 15–19.
 
-**How to resume:** Read AGENTS.md / CLAUDE.md / STATUS.md, open the next unchecked task in `docs/tasks/80`–`88`, stop at acceptance criteria.
+**How to resume:** Phase 9 P1 batch complete. Next work is P2/P3 Gitea backlog or homelab LLM wiring (`LLM_BASE_URL` in homelabstack `.env`).
 
 ## Development login/testing notes