- Production now runs from ~/freezmo-prod/current (release symlink) as a
standalone server on :3207; releases are exported by scripts/deploy.sh,
built and e2e-verified on a temp port, switched atomically, health-
checked via the new /api/healthz (commit + build time) and rolled back
automatically on failure. Last 5 releases kept; scripts/rollback.sh for
manual rollback. Shared env + data live in ~/freezmo-prod/shared/ and
are symlinked into every release (existing sync/push data migrated and
verified identical).
- Dev runs from ~/freezmo via freezmo-dev.service: next dev on :3217
(LAN only), its own build folder (.next-dev via NEXT_DIST_DIR), its own
data (~/freezmo/.freezmo-dev), DEV badge, dev favicon/theme colour,
"Freezmo Dev" PWA identity and never the production service worker.
- /api/healthz + the Settings Details version row (short commit + build
time). output: standalone in next.config (next start still works).
- AGENTS.md documents the workflow rules: only dev updates during a task,
temporary dirs/ports for test builds, and deploy-at-the-end as the only
production path. README gains an Environments and deploy section.
- Fix found by the rehearsal: lib/dev-mode.ts must not be a client module
(a server component importing a constant from a "use client" module
gets a truthy proxy, which baked dev metadata into production builds).
- SW -> freezmo-v34.
Entry points (no sixth tab): Ask in the Library header (desktop button,
phone icon), Ask about this deck in the deck menu, Ask a question on
study/homework results (compact result travels as context), and a Chat
row in the desktop sidebar. Phones get a full-screen route with back +
New chat; the tab bar steps aside.
- Client-side retrieval: chunk scoped decks (~800 tokens), BM25 against
the question + recent turns, top chunks under ~7k tokens, each
labelled with deck + section. Notes come from the deck's stored
sourceText (already kept for Regenerate); decks without it fall back
to generated content and the UI says so.
- Stateless server: /api/chat validates against the CHAT_MODELS
allowlist, trims the history to ~6k tokens, caps context at ~9k and
streams token/thinking/done/error NDJSON. /api/chat/extract reduces
attachments to text through the existing vision step.
- UI: user bubbles + full-width assistant text, streaming with stop,
auto-scroll that pauses with a jump-to-bottom button, safe markdown
(React nodes, no HTML injection; KaTeX dynamically imported only when
a message has math), Thinking row for reasoning, copy/regenerate/
edit-and-resend, model + context pickers (last model remembered),
composer with visualViewport keyboard handling, safe-area, 16px
textarea, Enter-to-send on desktop.
- Usage-limit failures keep the draft and show the shared message with
Settings/Ollama actions; stopping maps the reader abort to a clean
stopped state; offline disables sending with a reason.
- Settings: Keep notes text for chat toggle; deck menu: Delete notes
text. RATE_LIMIT_CHAT + CHAT_MODELS in .env.example and README.
- Tests: chat suite (63 checks, stub Ollama) wired into run-all.
Full set: 12/12 suites, 487 checks. SW -> freezmo-v32.
Also: tests/model-compare.mjs (diagnostic) benchmarks two models on the
same notes (latency, repair rounds, activity types) so the gpt-oss vs
gemma default can be decided on numbers; defaults unchanged.
Check my work (third homework mode)
- One model call with its own system prompt marks the student's own
attempt: verdict, what's right, each issue with why + fix, and the
corrected version that keeps everything the student got right.
Unasked-for extras are never errors. The attempt is typed/pasted
(photos of handwriting are documented as unsupported) and stored in
the record so Redo re-checks the same work.
- Server: mode "check" + solution on POST /api/homework (zod, 400
server.needSolution without one) and the same NDJSON stream; new
checkHomework() pipeline with repair-once validation.
Follow-ups
- New POST /api/homework/refine (plain JSON, RATE_LIMIT_REFINE, zod,
auth before body): shorter / more detail / explain, plain text back,
language preserved.
- Shorter + More detail + Original rewrite the finished answer and the
corrected version in place (FollowUpBar); every guide step has
Explain step N differently; failed calls say so without losing text.
Turn into a study deck
- Any result hands its content (answer, corrected version or assembled
guide) to New deck through a one-shot sessionStorage seed with the
options step pre-filled.
UI
- Phone step 2 gains the third card: tapping it reveals the solution
field + submit; desktop has three cards and the solution field shows
with the mode. Homework history labels and share/card previews cover
the new mode.
Verified: new phase-3 suite 20/20 against a stub Ollama (check payload,
missing-solution 400, refine validation, phone + desktop check flows,
shorter follow-up with restore, guide explain, deck hand-off) plus
regression 74/74, mobile 52/52, declutter 116/116, wizard 80/80,
animation 8/8, native 35/35, phase1 38/38, phase2 43/43, i18n 25/25.
Key parity 734/734. Lint + build clean. SW freezmo-v27.
Answer history and schedules
- Every answer is recorded (deck, activity, topic, correct/wrong, time)
in a versioned IndexedDB log (freezmo-learning v1, stores answers +
srs), mirrored in memory for synchronous React reads and capped at
20k answers (oldest pruned). StudyFeed records as it grades.
- Light SM-2 schedule per activity: 1 day, 3 days, then interval x
ease; wrong answers return in ten minutes with the ease nudged down.
Review, mistakes, stats
- Library 'Smart review' card: due count, mistakes count (latest
answer wrong — clears itself when answered correctly), today's goal
progress. Review sessions freeze their item list at start.
- New /review (due across decks, ?mistakes=1 global or ?deck= per
deck) built on a shared SessionFeed with the study feed's snap/keyboard
mechanics, wake lock and a results slide.
- New /stats: day streak, five-week calendar, accuracy per subject and
per deck, weakest topics (>=3 answers), daily-goal progress.
- Daily goal configurable in Settings (10/20/30/50).
Exam mode
- New /exam: pick decks, question count, optional timer; no hints
(a NoHints context hides hint rows and the self-check escape);
results with score, per-deck and per-topic breakdown; wrong answers
land in the mistakes set through the normal history.
Reminders (conditional)
- VAPID-gated: GET/POST/DELETE /api/push and POST /api/push/test with
the standard route pattern, RATE_LIMIT_PUSH, zod, auth before body.
Subscriptions stored server-side as endpoint + keys only (PUSH_DIR).
UI hidden entirely without keys; iOS install note; service worker
push/notificationclick handlers; scripts/send-reminders.mjs for a
cron/systemd timer; web-push added as the one server dependency.
Verified: new phase-2 suite 42/42 (history/SRS shapes from a real study
run, mistakes clearing, review session, stats math, exam timer and
breakdowns, VAPID-less hidden state plus the configured
subscribe/validate/test/delete path against a second instance) plus
regression 74/74, mobile 52/52, wizard 80/80, declutter 116/116,
animation 8/8, native 35/35, phase1 38/38, i18n 25/25. Key parity
709/709. Lint + build clean. SW freezmo-v26.
Safety net and errors, mobile-first, desktop unchanged in behaviour.
Backup and persistence
- Settings -> Data & backup: 'Export all data' writes one zod-validated
JSON (decks, folders, homework, progress, safe settings; the API key
is not part of the shape and never appears in the file).
- 'Import backup' validates, shows a preview with added/replaced
counts and offers merge or replace; invalid files get a friendly
error. Storage persistence (navigator.storage.persist) is shown and
requestable, with the result reported.
- A dismissible nudge appears after 5 new decks or 14 days without a
backup and stays snoozed for 14 days.
Opt-in encrypted sync (off by default)
- New lib/sync.ts: PBKDF2-SHA256 (310k) -> AES-GCM-256, derived key
kept as a non-extractable CryptoKey in IndexedDB (lib/sync-key.ts);
the passphrase and key never leave the browser.
- New POST/GET/DELETE /api/sync following the route pattern (locale,
cross-origin guard, body cap, rate limit RATE_LIMIT_SYNC, auth
before body parsing, zod validation). One atomic JSON file per
random 128-bit sync ID under SYNC_DIR; server-stamped updatedAt is
the conflict clock; SYNC_MAX_BYTES and SYNC_TOTAL_BYTES enforced
(413 / 507). The server only ever sees an opaque envelope.
- Settings -> Sync: enable with a passphrase, copy the sync ID, sync
now, last-synced time, explicit keep-this-device / use-synced-copy
conflict choice, join from another device (ID + passphrase, wrong
passphrase rejected), unlock after the key is gone, delete the
synced copy (works even on a locked device). Every data store marks
sync dirty on write; pulled payloads are applied without marking
dirty again.
Monthly usage vs short-term throttling
- lib/ollama.ts classifies the real 429 body (monthly/quota/upgrade
vs burst) into new 'usage_limit' and 'throttled' kinds, mapped in
every route and in the health probe.
- The client renders the same localized message in the deck flow,
homework, grading and the Settings banner: the required
upgrade/new-key copy with 'Open Settings' (jumps to and focuses the
key field) and 'Open Ollama'. Input and wizard drafts are kept so
Retry works right after swapping the key. No model fallback.
Verified: new phase-1 suite 38/38 (backup content + no key, import
preview/merge/replace, junk file, persistence row, nudge + snooze,
sync enable/upload/opaque blob/push/join/wrong passphrase/delete,
server validation 400/401, usage-limit copy and key focus, and the
real route mapping against a stub Ollama for both 429 variants) plus
regression 74/74, mobile 52/52, wizard 80/80, declutter 116/116,
animation 8/8, native 35/35, i18n 25/25. Key parity 638/638. Lint +
build clean. SW freezmo-v25.
Homework (/homework — entry points in the drawer and on the create screen)
- Assignment intake reuses the study pipeline: pasted text, photos
(camera/picker, compressed in the browser) and documents (PDF/DOCX/PPTX/
ODT/text) through lib/image + lib/documents, the same vision step and the
same language guard. Raw files never leave the device — only text and JPEG
data URLs reach the server.
- Two large choice cards: “Just make it” returns the finished homework
(answer only, plain text, copy + make it again); “Explain how to make this
homework” builds an interactive guide.
- The guide is one step per screen in the study-feed snap layout:
explanation, three progressive hints (nudge -> hint -> the answer) and an
optional checkpoint rendered by the existing activity components (graded
through /api/grade). The last screen assembles the solution from the step
answers, with a copy button. Checkpoints carry xp 0 so the guide stays free
of XP/streak chrome.
- Results live in freezmo.homework.v1 with a small history (open/delete);
“Make it again” re-runs from the stored assignment text (the vision
transcription when photos were used) — no original files are stored.
- The answer and the guide are written in the assignment's language, never
translated into the UI language.
Server
- POST /api/homework: NDJSON stage events like /api/generate, the same
http-guard, auth checked before parsing, and its own RATE_LIMIT_HOMEWORK
(documented in .env.example + the README table).
- New prompts in lib/prompts.ts and zod schemas in lib/schemas.ts with the
same single repair retry; TEXT_MODEL for content, VISION_MODEL only for
photos. New shared client pieces: useNotesIntake, lib/homework store,
lib/clipboard, and a shared NDJSON reader in lib/client-api.
Also finishing the mobile polish pass from the previous request
- Phone ice scene: the top-left orb moved behind the glass header (it sat
behind the page heading/subtitle). The settings sheet's scroll cue is now a
bare chevron inside a content fade (mask) instead of a chip over the status
banner, and the API key card's note/disclosure/actions are tightened.
- theme-color now follows the in-app theme (SSR metas still fall back to the
OS preference); apple-mobile-web-app-status-bar-style stays “default”
because iOS fixes the status-bar text colour at launch.
Verified: 78/78 homework checks (text/photo/PDF, both modes, EN+CS, dark+light,
360/390px, copy, history persistence, storage privacy, API language, axe), plus
the existing suites re-run (mobile 47/47, pixel contrast 0 below AA, i18n 27/27,
desktop regression 72/72). Lint + build clean; service worker freezmo-v6.
English stays the default; Czech is a full translation. Only the app chrome
is translated - notes and generated decks keep their own language, which is a
core feature, not a setting.
i18n
- lib/i18n-dict.ts: flat en/cs dictionaries (441 keys each), translate() and
translatePlural() with real Czech plural rules (1 sada / 2-4 sady / 5+ sad).
English is the default and the fallback for a missing key.
- lib/i18n.ts: browser store (freezmo.locale.v1), useI18n() hook, and the
x-freezmo-locale header. The hook's t/plural are bound to the locale React
is rendering with, so the first client render matches the server HTML and
then re-renders in the stored language; module-level t is only for event
handlers and non-React helpers.
- <html lang> follows the choice (inline script + setLocale).
UI
- Settings -> Language segmented control (English / Cestina), plus an EN/CS
button on the key gate for first run.
- Every screen translated: sidebar, header, tab bar, library, deck cards,
folder bar, create flow, all seven activities, study chrome, results,
settings + health panel, API-key gate and help, install sheet, offline and
the deck-not-found page.
- Plural-aware counts everywhere (decks, folders, activities, runs, photos,
documents, pages, words, wrong attempts).
Server
- API routes and guards (generate, grade, health, http-guard, rate-limit,
ollama error mapping) answer in the request's language, so error cards and
the health panel never fall back to English mid-sentence.
Fixes
- A hydration mismatch when a non-default locale was stored silently dropped
the .dark class after hydration; render-time t() is now bound to the
hook's locale (UploadScreen.describeExtraction, ResultsScreen.answerFor,
DeckCard's folder fallback).
- Bumped the service worker cache to freezmo-v4 for this deploy.
Tests: i18n 27/27 (gate toggle, persistence, html lang, Czech plurals,
settings switch) and a rebuilt regression suite 72/72 (English default,
library, create, all 7 activity types, results, settings, install sheet,
offline, PWA, security, Czech smoke).
Next.js 15 app that turns photos, PDFs, DOCX/PPTX and text into a swipeable
feed of AI-generated activities (flashcards, multiple choice, true/false,
fill-in-the-blank, matching, ordering, short answer) using the user's own
Ollama models.
Highlights:
- All AI calls run in server route handlers; the browser never talks to Ollama.
- Bring-your-own-key mode (default): the key lives only in the browser and is
sent per-request; server, local-daemon and shared-key modes also supported.
- Client-side document extraction (pdf.js, mammoth, jszip) with scanned-PDF
page rendering for the vision model; notes stay in their original language.
- ChatGPT-style shell: collapsible sidebar / mobile drawer, neutral token-based
light + dark themes, folder organisation, PWA (manifest, SW, offline page).
- Strict JSON validation with zod + auto-repair, rate limiting, and request
guards (cross-origin, body-size, API-key sanitising).
- setup.sh bootstraps deps, .env.local and a production build.
Gitignore excludes .env*, the generated pdf.js worker and local scratch files.
