feat: BotE Calculator & Scale Sandbox + Smart SRS Knowledge Graph (v1.4.0)

[arvarik]

Summary

Implements two major features against their business specs, plus a critical SRS bug fix found during implementation.

🧮 Back-of-the-Envelope Calculator & Scale Sandbox (/calculator, ⌘8, quick modal ⌘E)

• Pure math engine (src/utils/bote.js) — standard interview conventions: QPS = DAU × req/user ÷ 86,400, storage with retention × replication × overhead, the 80/20 working-set cache, peak bandwidth, and server/shard counts from target CPU saturation. SI-unit formatters up to EB/Tbps, safe at zero and at 2B-DAU extremes.
• Interactive sandbox — log-scale sliders + numeric inputs, six scenario presets (URL Shortener, Video Streaming, E-Commerce Flash Sale, Chat, Social Feed, Ride Sharing), and every stat shows the formula that produced it with real numbers substituted.
• Numbers Every Engineer Should Know — canonical latency table (grounded in the Dean/Norvig numbers plus modern cross-AZ/cross-region RTTs) with click-to-apply Latency Budget composer, powers-of-two table, and handy constants.
• AI Sanity Checker — POST /api/calculator/audit returns a structured, severity-ranked critique (verdict, findings with concrete fixes, omitted real-world factors). Verified live: it correctly recomputed a deliberately wrong cache figure.
• Omnipresent — dedicated page plus a global modal reachable from Chat, Guide, and Builder; state persists across navigation (persisted Zustand store). 1-click export as Markdown with LaTeX to clipboard, Chat draft, or any Guide section.

🕸️ Smart SRS Knowledge Graph & Adaptive Prerequisite Engine (/graph, ⌘7)

• ~60-concept prerequisite DAG grounded in the standard learning progression (fundamentals → building blocks → distributed mechanics → paradigms), including the spec's chain Consistent Hashing → Distributed KV Stores → Virtual Nodes. Shared client/server module, validated acyclic at boot and in tests. Two topics added to the taxonomy: Caching Strategies & Invalidation and Consensus & Coordination.
• Live retention heatmap — 🟢 mastered (ease ≥ 2.5, interval > 21d), 🟡 learning/due (≤ 48h), 🔴 decayed/fragile, ⚪ unseen. Cards auto-link to concepts by keyword phrases with guide-topic fallback. Node colors refresh after every graded card without a reload.
• Exploration UX — deterministic force layout (prerequisites flow left→right), pan/zoom, hover lineage highlighting (prereqs in indigo, unlocked concepts in teal), search, pillar chips, readiness filters, and a slide-over panel deep-linking to the Guide blueprint, related whiteboards (/builder?board=<id>), and due flashcards with a 1-click node study session.
• Curated learning tracks — six tracks (Senior Distributed Systems, Storage & Consistency, …) that number their concepts in topological order on the canvas and build prerequisite-ordered study sessions.
• Adaptive remediation — rating an advanced card Again checks its direct prerequisite concepts for shaky cards (new/learning/lapsed/overdue/low-ease) and queues up to 3 into the next session via a remediation_queue table. They lead the queue with a Foundation Checkup banner. Solid foundations produce no remediation. Reviewing a queued card resolves it.

🐛 Fixed: silent card-grading failure

The client sent POST to the review endpoint while the server only registered PUT — every grade 404'd and was swallowed by the catch handler. The client now sends PUT; the server accepts both verbs (queued offline mutations may still carry POST). The SM-2 scheduler was extracted to server/srs/scheduler.js for reuse by the graph routes.

Testing

• 250 tests / 18 files, all passing — new suites cover formula precision & unit conversions (reference case + zero/extreme boundaries), graph integrity (unique ids, acyclic, topo order, track expansion closure), cycle prevention, card→node matching precedence, health bucket boundaries (ease 2.5 / interval 21 / 48h window), remediation planning (shaky detection, cap, weakest-first, one-hop-only), and UI interaction (calculator recalculation/presets/budget, graph filters/panel/tracks).
• Lint clean, production build clean.
• Verified end-to-end in Chrome against the running dev server: graph heatmap with seeded data across all four health states, node panel deep links, track sessions, foundation-checkup banner in a live interleaved session, quick modal in Chat, AI audit round-trip, and light theme.
• README + docs refreshed (api.md, user-guide.md, srs-algorithm.md, architecture.md, AGENTS.md) with new screenshots for all pages.

🤖 Generated with Claude Code

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.