Implementation patterns — how features are wired together using the stack.
Essential patterns for adding 3D to SvelteKit. Copy-paste starting points.
SvelteKit 2 + Svelte 5 patterns for integrating 3D experiences. Framework-first approach with progressive enhancement.
Final architectural recommendation for integrating AI tool-calling into the Desk workspace.
Hybrid retrieval combining knowledge graphs (Neo4j) with vector embeddings for more accurate, explainable AI responses.
"Harness" is a lens we use to audit the bot's post-prompt-dispatch machinery — not a module.
Two-layer retrieval: llmwiki is the primary answer surface; rawrag is the audit and drill-down trail.
Compact serialization format optimized for LLM input. Reduces tokens by 30-60% compared to JSON.
src/lib/server/analytics/hook.ts exports analyticsCollector but it is not wired into the sequence(...) in src/hooks.server.ts.
Persistent, conversational AI chat accessible via sidebar trigger or keyboard shortcut.
Rules for organizing components using the Atomic Design pattern: atoms → molecules → organisms → templates.
UI patterns for when there's no data to display. Critical for first-run experience and edge cases.
Central registry for all keyboard shortcuts in the app shell. Enables discoverability, conflict resolution, and accessibility.
The app shell layout, breakpoints, z-index layers, and accessibility requirements.
Visual feedback during data fetching and navigation. Critical for perceived performance and user confidence.
Progressive Disclosure with hierarchical navigation. Users land on pages with a single click; subpages/sections are revealed via dropdown.
In-app notification system with full-page notification center, preference management, and real-time updates.
Per-page header inside the main content area. Not a global header — each page optionally includes this component for its title and actions.
Global navigation and search via keyboard shortcut or sidebar trigger.
Managing authentication state within the app shell: session expiry, re-authentication, and graceful degradation.
User configuration for appearance, language, privacy, and accessibility. This page covers both:
Orchestration of state across app shell components: sidebar, modals, theme, notifications, and user session.
The sidebar is the primary navigation element, adapting responsively for desktop and mobile.
Ephemeral feedback messages that appear in response to user actions. Distinct from the notification center (persistent) - toasts are immediate, temporary, and action-specific.
User account management for identity, security, and GDPR compliance. High-stakes operations requiring careful UX design.
Background work in the application falls into three categories distinguished by trigger mechanism, not code structure.
The same backend logic must serve five client types: human UI (SvelteKit form actions and load functions), AI agents (Vercel AI SDK tool calls), external API (REST endpoints)…
How to manage database schema changes across development and production environments using Drizzle ORM.
Graph data model for relationships, navigation, and future RAG capabilities.
Cross-database reference integrity in a multi-store architecture.
Drizzle ORM schema for relational data and Better Auth integration.
REST endpoints with SvelteKit +server.ts files.
Passwordless session-based authentication using Better Auth with Drizzle ORM.
Tri-target deployment: Vercel (Node.js), Vercel (Bun experimental), and Koyeb (Bun container).
SvelteKit error handling patterns for pages, forms, and API routes.
Type-safe forms with real-time validation using Superforms and Valibot.
URL-based locale routing with compile-time translations using Paraglide JS v2.
SvelteKit hooks for request interception, authentication, and cross-cutting concerns.
Self-documenting template where each page tests the features it documents.
Implementation blueprint for the Progressive Revelation system.
Request rate limiting to prevent abuse, brute force attacks, and resource exhaustion.
Svelte 5 runes for explicit, portable reactivity.
How users connect external notification channels (Telegram, Discord) to their account.
Backend architecture for multi-channel notification delivery.
Database tables for multi-channel notification system.
Frontend architecture for channel connections and notification settings management.
The project has no test infrastructure yet. This document defines what to build, why, and how.