managarten/apps/mana/CLAUDE.md

CLAUDE.md — Mana Unified App

Project-level guidance for apps/mana/. For monorepo-wide patterns (auth, services, dev commands, env vars), see the root CLAUDE.md.

Project Overview

Mana is the unified web app at mana.how, serving 27+ product modules (todo, calendar, contacts, chat, notes, dreams, memoro, cards, picture, presi, music, storage, …) under one SvelteKit build, one IndexedDB, one auth session, one deployment.

apps/mana/apps/
├── web/        # SvelteKit 2 + Svelte 5 unified app — the main surface
└── landing/    # Astro static landing → Cloudflare Pages

Note: apps/mana/apps/mobile/ was removed on 2026-04-20 along with five other product mobile apps (cards, chat, context, picture, traces). The only remaining Expo/React Native surface in the repo is apps/memoro/ apps/mobile/.

Module System

Each module lives in apps/web/src/lib/modules/{name}/ and registers itself via module.config.ts. Module state is split into three files:

File	Role
collections.ts	Dexie table references + (sometimes) seed data
queries.ts	Read-side — Dexie liveQuery hooks, type converters, pure helpers for $derived
stores/*.svelte.ts	Write-side — mutation methods. Never reads for UI rendering (queries.ts does that). Only reads when a mutation needs existing state (toggle, increment).

Module store pattern

// modules/todo/stores/tasks.svelte.ts
export const tasksStore = {
  async createTask(input: {...}) {
    const newLocal: LocalTask = { ...input, id: crypto.randomUUID() };
    const plaintextSnapshot = toTask({ ...newLocal });
    await encryptRecord('tasks', newLocal);
    await taskTable.add(newLocal);
    return plaintextSnapshot;
  },
};

// modules/todo/queries.ts
export function useAllTasks() {
  return useLiveQueryWithDefault(async () => {
    const locals = await db.table<LocalTask>('tasks').orderBy('order').toArray();
    const visible = locals.filter((t) => !t.deletedAt);
    const decrypted = await decryptRecords('tasks', visible);
    return decrypted.map(toTask);
  }, [] as Task[]);
}

Data Layer (Local-First)

The app reads and writes IndexedDB first, then syncs to mana-sync (Go, port 3050) in the background. One Dexie database (mana) holds 120+ collections from every module — colliding table names get a module prefix (e.g. todoProjects, cardDecks, presiDecks).

User action (e.g. tasksStore.createTask)
        │
        ▼
Module store builds the LocalRecord
        │
        ▼
encryptRecord(tableName, record)
        │
        ▼
table.add(encryptedRecord)            ← Dexie write
        │
        ▼
Dexie hooks (database.ts):
  - stamp userId
  - stamp __fieldTimestamps per field
  - stamp __lastActor + __fieldActors (user / ai / system — see AI Workbench)
  - record into _pendingChanges  (tagged with appId + actor)
  - record into _activity
        │
        ▼
Sync engine (sync.ts) — debounced 1s
  - groups changes by appId
  - POSTs to mana-sync
        │
        ▼
mana-sync → PostgreSQL with field-level LWW + RLS
        │
        ▼
Other clients pull via SSE / polling
        │
        ▼
applyServerChanges → Dexie hooks (suppressed) → liveQuery → decryptRecord → UI

Deep dive: apps/web/src/lib/data/DATA_LAYER_AUDIT.md — sync engine, retry/backoff, quota recovery, telemetry, RLS, encryption rollout, threat model. Single most important file for understanding how the app works under the hood.

At-Rest Encryption

User-typed content in 27 tables is encrypted with AES-GCM-256 before it touches IndexedDB. Master key lives in mana-auth (KEK-wrapped) and is fetched on login.

Mode	Default	What Mana can decrypt
Standard	✅ Yes	The user's master key, via the server-side KEK
Zero-Knowledge	Opt-in (Settings → Sicherheit)	Nothing — recovery code lives only with the user

When writing module code that touches sensitive fields:

1. Add the table to apps/web/src/lib/data/crypto/registry.ts with the field allowlist
2. await encryptRecord(tableName, record) before table.add() / table.update()
3. await decryptRecords(tableName, visible) after the Dexie query, before the type converter
4. The Dexie hook in database.ts does NOT auto-encrypt — every store does it explicitly. This is by design (Web Crypto is async, hooks are sync).

Defaults: encrypt for new user-typed text fields; plaintext for IDs / timestamps / sort keys / enum discriminators.

User-facing docs: apps/docs/src/content/docs/architecture/security.mdx.

Routing

apps/web/src/routes/
├── (auth)/              # Public auth pages (login, register, recovery)
├── (app)/               # Auth-gated app surface — 27+ module routes
│   ├── dashboard/       # Customizable widget grid
│   ├── settings/
│   │   └── security/    # Vault status + recovery code + ZK opt-in
│   ├── todo/            # …and many more module routes
│   └── …
└── api/                 # SvelteKit API endpoints (rare; most data is local-first)

The (app) group is wrapped by AuthGate, which redirects unauthenticated users to /login and reads the access tier from the JWT to gate beta/alpha/founder-only modules.

Legacy Supabase: removed. Anything mentioning @supabase/ssr, safeGetSession(), or event.locals.supabase is leftover from a much earlier iteration and should be deleted on sight.

Path Aliases (apps/web/svelte.config.js)

$lib → src/lib · $components → src/lib/components · $stores → src/lib/stores · $utils → src/lib/utils · $types → src/lib/types · $server → src/lib/server

Auth Access Pattern

Auth state lives in $lib/stores/auth.svelte.ts. The current user id is also pushed into $lib/data/current-user.ts so the Dexie creating-hook can auto-stamp userId on every record. Module stores never need to know who the current user is — they just write, and the hook stamps the right userId.

Development Commands

For full local-dev (Mana Auth + mana-sync + web together), use the root-level pnpm run mana:dev or pnpm dev:*:full commands. See root CLAUDE.md and docs/LOCAL_DEVELOPMENT.md.

Web-app-only:

cd apps/mana/apps/web
pnpm dev          # Dev server on :5173
pnpm build        # Production build
pnpm preview      # Preview production build
pnpm check        # svelte-check type check
pnpm lint         # Format check + ESLint
pnpm format       # Prettier write
pnpm test         # Vitest (unit + integration with fake-indexeddb)
pnpm test:e2e     # Playwright

Tech Stack

• Web: SvelteKit 2 + Svelte 5 (runes mode), TailwindCSS, Vite
• Auth: Mana Auth (Better Auth + EdDSA JWT) via @mana/shared-auth
• Data: Dexie.js (local-first) + mana-sync (Go) backend
• Encryption: AES-GCM-256 via Web Crypto, server-wrapped MK with optional zero-knowledge
• Local AI: @mana/local-llm (Gemma 4 E2B, WebGPU) + @mana/local-stt (Whisper, WebGPU) — both run entirely in-browser via transformers.js
• Testing: Vitest, Playwright
• Mobile: removed (see note at top) — Expo stack lives only in apps/memoro/apps/mobile/ now

Svelte 5 runes are mandatory — no legacy let count = 0; $: doubled = count * 2. Always $state, $derived, $effect. See .claude/guidelines/sveltekit-web.md.

Scene Scope

Each workbench scene can carry scopeTagIds — a per-scene tag filter that module queries honour via filterBySceneScopeBatch from $lib/stores/scene-scope.svelte. When the filter hides everything, users need to see why.

When a module wires the scope filter, wire the empty state too:

<script lang="ts">
  import ScopeEmptyState from '$lib/components/workbench/ScopeEmptyState.svelte';
  import { hasActiveSceneScope } from '$lib/stores/scene-scope.svelte';
</script>

{#if items.length === 0}
  {#if hasActiveSceneScope()}
    <ScopeEmptyState label="Aufgaben" />
  {:else}
    <p class="empty">Noch keine Aufgaben</p>
  {/if}
{/if}

ScopeEmptyState renders a subdued "Bereichsfilter verbergen alles" message plus a one-click "Bereich zurücksetzen" button that calls workbenchScenesStore.setSceneScopeTags(activeSceneId, undefined). SceneAppBar already shows a Funnel badge on scoped scene pills; the module doesn't need to duplicate that signal. Plan: docs/plans/scene-scope-empty-state.md.

AI Workbench

The companion is a second actor that works alongside the human in every module. Full pipeline live end-to-end:

• Actor attribution — every event, record, and sync row carries { kind, principalId, displayName } (+ mission/iteration/rationale for AI). principalId is the userId / agentId / system:<source> sentinel; displayName is cached at write time so rename doesn't rewrite history. Factories in @mana/shared-ai/src/actor.ts; runtime ambient context in src/lib/data/events/actor.ts.
• Agents — named AI personas that own Missions. /ai-agents module for CRUD (policy editor, memory, budget, concurrency). Default "Mana" agent auto-bootstrapped on first login; legacy missions backfilled. data/ai/agents/{store,queries,bootstrap}.ts.
• AI policy — per-tool auto | propose | deny. Lives on the agent (agent.policy). Proposable tool names come from @mana/shared-ai's AI_PROPOSABLE_TOOL_NAMES; the mana-ai service runs a boot-time drift guard against the same list. Resolution in src/lib/data/ai/policy.ts; executor loads agent.policy for every AI write.
• Proposal inbox — drop <AiProposalInbox module="…" /> into any module page to render pending proposals inline with approve / freitext-reject buttons. Cards show the owning agent's name + avatar chip. Wired in /todo, /calendar, /places, /drink, /food, /news, /notes. The mission-detail view also embeds a cross-module inbox (<AiProposalInbox missionId={id} />): shows all pending proposals for that mission across all modules with a module-badge per card, so the user can review and approve without navigating to individual module pages.
• Reasoning loop — the foreground Runner chains up to 5 planner calls per iteration. Read-only tools (list_notes, get_task_stats, etc.) execute inline as auto-policy, their outputs are fed back as synthetic ResolvedInputs for the next planner call. The loop exits when a propose-policy tool is staged (human must approve), the planner returns 0 steps, or the budget exhausts. This enables "read → reason → act" missions like "list all notes and tag them" in a single run. Code: data/ai/missions/runner.ts reasoning loop.
• Missions — long-lived autonomous work items at /ai-missions with concept + objective + linked inputs + cadence + owning agent (AgentPicker in the create flow). Both the foreground tick AND the server-side mana-ai service produce plans under the agent's identity; data/ai/missions/server-iteration-staging.ts translates server-source iterations into local Proposals on sync.
• Input picker — <MissionInputPicker> sources candidates from the input-index registry (notes / kontext / goals / tasks / calendar). The Runner resolves via the parallel input-resolvers registry. Encrypted tables (notes, tasks, …) decrypt client-side only.
• Auto-injected context — the Runner automatically appends the user's kontextDoc singleton (decrypted client-side) to every planner call as a standing-context input, unless already linked manually. For missions whose objective matches research keywords (recherchier|research|news|…), a web-research pre-step runs the news-research RSS pipeline (discoverByQuery + searchFeeds) and injects results with explicit save_news_article instructions.
• Debug log — per-iteration capture of system/user prompts, raw LLM responses, resolved inputs, and auto-tool outputs. Stored in local-only Dexie table _aiDebugLog (never synced — contains decrypted user content). Toggled via localStorage('mana.ai.debug') (on by default in DEV). Rendered as expandable <AiDebugBlock> under each iteration card with copy-as-JSON button. Code: data/ai/missions/debug.ts, components/ai/AiDebugBlock.svelte.
• Scene lens — workbench scenes can bind to an agent via scene.viewingAsAgentId (context menu → "An Agent binden…"). Pure UI lens, not a data-scope change. SceneAppBar shows the agent avatar on bound scene tabs.
• Workbench timeline — /ai-workbench renders every AI-attributed event grouped by mission iteration with per-agent filter, per-module, per-mission. Each bucket header shows agent avatar + name + mission title. Per-bucket Revert button undoes the iteration's writes via data/ai/revert/ (TaskCreated → delete, TaskCompleted → uncomplete, etc., newest-first). Separate "Datenzugriff" tab exposes the server-side decrypt audit (for missions with Key-Grants).

Tool Coverage (59 tools, 19 modules)

Agents interact with the app through tools — each one either auto (executes silently during reasoning) or propose (creates a Proposal card the user must approve). Source of truth: AI_TOOL_CATALOG in @mana/shared-ai/src/tools/schemas.ts — both webapp policy (src/lib/data/ai/policy.ts) and server-side planner (services/mana-ai/src/planner/tools.ts) derive from it automatically, so drift is structurally impossible.

Module	Propose	Auto
todo	create_task, complete_task, complete_tasks_by_title	get_task_stats, list_tasks
calendar	create_event	get_todays_events
notes	create_note, update_note, append_to_note, add_tag_to_note	list_notes
places	create_place, visit_place	get_places, get_current_location
drink	undo_drink	get_drink_progress, log_drink
food	—	nutrition_summary, log_meal
news	save_news_article	—
news-research	research_news	—
journal	create_journal_entry	—
habits	create_habit, log_habit	get_habits
contacts	create_contact	get_contacts
quiz	create_quiz, update_quiz, add_quiz_question, update_quiz_question, delete_quiz_question	list_quizzes, get_quiz_questions, get_quiz_stats
goals	create_goal, pause_goal, resume_goal, complete_goal	list_goals, get_goal_progress
mood	log_mood	get_mood_today, get_mood_insights
myday	—	get_myday_summary
events	suggest_event	discover_events
finance	add_transaction	get_month_summary, list_transactions
times	start_timer, stop_timer	get_time_stats, get_timer_status, list_projects
wetter	—	get_weather, get_rain_forecast

Server-side web-research: mana-ai calls mana-api's /api/v1/news-research/discover + /search directly before the planner prompt is built (pre-planning injection). Missions with research-keyword objectives get real article URLs + excerpts injected as a synthetic ResolvedInput. See services/mana-ai/src/planner/news-research-client.ts.

Templates

Pre-configured starter-kits at /agents/templates — two sections:

• Agent-Templates (with AI): Recherche-Agent, Kontext-Agent, Today-Agent
• Workbench-Templates (no AI): Calmness, Fitness, Deep Work

Each template bundles: optional agent + optional scene layout + optional starter missions (paused) + optional per-module seeds. Template shape: WorkbenchTemplate in @mana/shared-ai/src/agents/templates/types.ts. Applicator: src/lib/data/ai/agents/apply-template.ts. Seed-handler registry: src/lib/data/ai/agents/seed-registry.ts — modules register via side-effect imports in missions/setup.ts. Current handlers: meditate, habits, goals. Plan: docs/plans/workbench-templates.md.

Full architecture (Planner prompt + parser in @mana/shared-ai, server-side runner, Postgres actor column, materialized snapshots, Multi-Agent gating, server-side web-research, Prometheus metrics + status.mana.how integration): docs/architecture/COMPANION_BRAIN_ARCHITECTURE.md §20 (AI Workbench) + §21 (Mission Grants) + §22 (Multi-Agent Workbench).

Reference Documents

Path	Purpose
apps/web/src/lib/data/DATA_LAYER_AUDIT.md	Data-layer + sync deep dive, encryption rollout, threat model, Actor attribution, backlog
docs/architecture/COMPANION_BRAIN_ARCHITECTURE.md	Companion brain + AI Workbench (Actor, Policy, Proposals, Missions roadmap)
apps/docs/src/content/docs/architecture/security.mdx	User-facing security walkthrough
apps/docs/src/content/docs/architecture/authentication.mdx	Auth flow + JWT structure
Root CLAUDE.md	Monorepo overview, services, dev commands, env vars