- **Electron as the product runtime** — our product UI remains the shared Next.js web app; `apps/desktop` and `apps/packaged` provide a thin Electron host around the same daemon/web sidecars.
- **What it is:** Open-source personal AI assistant with many integrations, auto-fetch into a memory tree, local editable knowledge, and a TokenJuice-style compression layer before LLM use.
- **Why it matters:** It is the clearest reference for "connect sources once, then the agent wakes up with compressed context already available."
- **What we borrow:**
- Connector-driven ingestion as a first-class loop, not a prompt the user rewrites each time.
- Editable local memory tree rather than opaque vector-only recall.
- Token compression as an optional stage before agent context injection.
- **What we don't:**
- General personal-assistant scope, messaging/voice/meeting participation, and bundled subscription/model routing.
- **What it is:** Self-improving agent with persistent memory, skills created from experience, skill improvement during use, scheduled automations, and a large skill hub.
- **Why it matters:** It turns "agent learns from work" into a product loop instead of a side effect.
- **What it is:** Minimal self-evolving autonomous agent framework that crystallizes solved tasks into a personal skill tree for direct reuse.
- **Why it matters:** It names the core loop OD needs for design work: solve once, verify, save the execution path, and recall it with less context next time.
- **What we borrow:**
- Skill crystallization from successful tasks.
- Layered memory and direct recall to reduce prompt size.
- A bias toward small composable primitives instead of a heavy agent framework.
- **What we don't:**
- Broad uncontrolled desktop authority. OD keeps daemon, connector, filesystem, and review gates explicit.
- **What it is:** Tauri desktop app for managing five CLI tools (Claude Code, Codex, Gemini CLI, OpenCode, OpenClaw). Provider management, MCP server config, skills install, session browsing. SQLite at `~/.cc-switch/cc-switch.db`. **Skills dir at `~/.cc-switch/skills/` with symlinks into each agent's config dir.** 50+ provider presets.
- **Why it matters:** Shows exactly how to live beside multiple code-agent CLIs without stepping on their config.
- Knowledge of per-agent config dir locations (`~/.claude/`, `~/.codex/`, …).
- "Provider presets" idea — a curated list we can ship so users don't have to hand-enter endpoint URLs for OpenAI-compatible relays.
- **What we don't:**
- Tauri / desktop app — not our shape.
- Provider-switching as core feature — we defer that to the underlying agent. If a user wants to switch providers inside Claude Code, they use Claude Code's config, not ours.
- **What it is:** A Claude Code skill producing magazine-style, horizontal-swipe web decks. Structure: `SKILL.md` + `assets/template.html` + `references/{components,layouts,themes,checklist}.md`. 6-step workflow. Single-file HTML output with embedded CSS/WebGL. Keyboard/scroll/touch navigation.
- The workflow and assets are bundled under [`design-templates/guizang-ppt/`](../design-templates/guizang-ppt/) with the upstream license preserved and OpenDesign metadata/file-handoff integration applied.
- **What we don't:** We do not require the retired `od skill add` flow; the rendering template ships in the catalogue and can still be consumed as a portable `SKILL.md` bundle.
| [Cursor rules](https://docs.cursor.com/) | Historical format inspiration only; the current Cursor Agent adapter receives OpenDesign's composed prompt over stdin and uses the same CWD-local resource staging as other runtimes |
| [Shadcn/ui](https://ui.shadcn.com/) | Component and registry reference; the product shell now uses repository-owned primitives from `@open-design/components` |
| [Vercel AI SDK](https://sdk.vercel.ai/) | Historical API-fallback exploration; current BYOK providers run through the OpenCode adapter rather than a direct SDK adapter |
| [Puppeteer](https://pptr.dev/) | Reference for template/skill-owned browser rendering; core desktop PDF export uses Electron `printToPDF` or the screenshot/PDF pipeline |
- **SQLite for artifacts** (from [Open CoDesign][ocod] and [cc-switch][ccsw]). Plain files + JSONL history are reviewable in git, trivially portable, and match the "skills are files" ethos.
- **Bundled model router** ([`pi-ai`][piai] from [Open CoDesign][ocod]). The user's code agent already routes. Two routers is worse than one.
- **PostgreSQL + pgvector** (from [multica][multica]). We don't embed anything in MVP. When we do, SQLite + `sqlite-vec` is enough for single-user scale.
- **Board / issue model** (from [multica][multica]). Off-brand for a design tool.
This file is maintained. When we add an adapter or borrow a pattern from a new upstream, add it here with the same three-question format. When upstream licensing or direction changes materially, flag it here and cross-link from [`spec.md`](spec.md).
