"The model selection in teach it back is kinda broken at least on desktop":
the desktop panels were position:absolute children of whatever container held
their chip. In chat that is the composer at the bottom of the column, which
works, but Teach it back keeps its chips inside a scrolling form card — so the
panel opened upward out of the card and its top rows (including the selected
model) were cut off. The probe measured top: -66px and no way to reach the
first rows by scrolling the list.
- New components/chat/PickerPopover.tsx: the desktop surface for both pickers,
portalled to <body> and position:fixed, anchored to the opening chip
([data-picker-trigger][aria-expanded="true"]), opening on the side with room
and capped to the viewport so the panel scrolls as a whole with its heading
and note attached. It closes on an outside press, Escape, resize or a page
scroll — but not on scrolling its own list.
- ContextPicker and ModelPicker use it on desktop; the phone sheets and their
inner list caps are unchanged (the list max-height became a parameter: the
sheet caps it, the desktop panel scrolls instead).
- teach-test pins it: the panel's rect is fully inside the viewport,
elementFromPoint over it lands on the panel, it is not a descendant of the
scrolling card, and picking a model updates the chip + freezmo.chat.model.v1
(46 checks). Chat and the phone sheets re-verified unchanged (110 + 128).
Version 0.8.3, SW v105.
A new mode where the user explains a topic in their own words and the AI
plays a curious classmate: short reactions, exactly one follow-up question
at a time, never "correct"/"wrong", never lecturing, and everything it knows
comes from the user's own decks and notes.
- Entry points: a sidebar row next to Exam mode and a "Teach it back" action
in every deck's menu (which prefills the topic and scope).
- Start screen: deck or typed topic, friendly/skeptical student, the same
context and model chips as Chat (/teach?deck=<id> pre-selects the deck).
- Session: the student prompt is a mode of the existing /api/chat route, so
model resolution, allowlists, rate limits and the streaming client are the
same code; the personality lives in lib/prompts.ts (TEACH_*). The deck's
cards travel as a numbered checklist and the student is told to ask about
the ideas that were skipped; it must say so rather than guess when the
notes cover nothing. One constant (TEACH_MAX_EXCHANGES = 10) ends the
session, or the Finish button does.
- Recap: a teach_summary call (JSON, one repair retry) yields well / gaps
with card references / next steps; each gap links to the card it means.
"Add missed cards to review" schedules those activities due now; "Turn
gaps into cards" generates one flashcard per gap from the deck's notes and
shows an editable review list before saving.
- Sessions are chat conversations with a mode:"teach" marker + session
record: same IndexedDB history, backup and sync, filtered out of the Chat
screens; the recent list reopens a recap or continues a session.
- Reused pieces: MessageRow, Composer (now optional attach/scope chips and a
mic), Markdown/KaTeX, AiDisclaimer, ContextPicker/ModelPicker, plus a
shared useChatModels hook and useKeyboardInset extracted from Chat.
- Dictation: Web Speech API recognition (lib/speech-input.ts); the mic only
renders where the browser supports it, and dictated phrases land in the
draft.
- All strings in en + cs. teach-test: 44 checks (API prompts and guards,
both entry points, the session, recap actions, history, exchange limit,
360/390px layouts, mic on/off). Version 0.8.1, SW v103.
