The generated videos weren't good enough — pull the films card, its
styles, and the video assets. The service-worker hardening stays: media
paths bypass the cache (Range/206 responses) and only full 200 responses
are cached.
Co-Authored-By: Claude Fable 5 <[email protected]>
Two short macro films generated on the local ComfyUI instance (Wan 2.2
14B t2v + lightx2v 4-step LoRAs, 9 clips at 832x480) and stitched with
crossfades and burned-in phase labels: roast-physics.mp4 (charge, drying,
Maillard, first crack, drop) and brew-physics.mp4 (grind, bloom,
immersion, espresso). A "Films — the physics in motion" card on the
Academy menu plays them with posters; opening a lesson hides and pauses
them. The service worker now bypasses /academy-video/ and /academy-audio/
(Range/206 responses the Cache API rejects) and only caches full 200s.
Co-Authored-By: Claude Fable 5 <[email protected]>
The planner's Artisan .alog drawer now shows what's attached with a
"Remove reference curve" control (re-rendered per drawer open), so adding
a reference curve is no longer a one-way door; /api/alog shares the 8 MB
body cap so real-sized logs parse instead of failing with "bad_request".
Deep-evaluation fixes: editing a brew of an archived bean no longer
silently detaches the bean; the roasts pending-review poll no longer
wipes in-progress after-roast edits; roasters gain an Edit (rename/model)
action; the gear page refuses to autosave over a failed load; duplicating
a plan carries its custom name; cupping sessions can attach a plan after
creation (ownership-checked PUT + selector); admin user deletion also
refreshes plans/audit; cupping cup-count subtitle stays live; roasts
error-row colspan corrected. Regression tests cover the new cupping PUT
and the /api/alog body cap.
Academy scenes drop the flat paper-cutout look: shared defs provide
radial-gradient shading on every bean/half-bean/particle, flame gradients
with radiant halos, soft ground shadows, and a warm-lit stage background;
fill-shift animations now ride a partial-opacity tint overlay so shading
survives the color change.
Co-Authored-By: Claude Fable 5 <[email protected]>
- New flavor-wheel.js: two-ring sunburst (families inner, descriptors
outer) in SCA hue families; tapping a wedge toggles the same flavor
tag as the checkbox list; selections render bold with a dark outline;
respects the tag limit; labels drawn above all wedges so long names
never clip (screenshot-verified in headless Chromium)
- Cupping radar: one hue per attribute on labels + vertex dots
Co-Authored-By: Claude Fable 5 <[email protected]>
Root cause of the broken diagrams found and fixed: CSS transform
animations were overriding the SVG positioning transform on the same
element, throwing animated elements (beans, flames) to the canvas
origin — the bean/flame helpers now split position (outer group) from
animation (inner group).
Every scene was then rendered in headless Chromium and screenshot-
inspected, with fixes applied and re-verified: drying-scene label
collision, RoR ideal-line label overlap, first-crack dot off the curve,
composition bar stagger orphaning its labels, moka flame hidden behind
the base.
New opening slide (user request): 'The whole journey, in one bean' — a
large bean on a 14-second loop living the full roast (heat in, drying
steam out, green→yellow→brown, pressure swell, crack) synced to a
stage timeline with a riding marker; narrated like the rest.
Co-Authored-By: Claude Fable 5 <[email protected]>
- Roasting is now a single continuous 11-slide lesson ('The roast,
start to finish') merging the six lessons' content; brewing keeps its
four per-family lessons
- Progress tracking: resume-where-you-left-off per lesson, progress bar
and completed badge on the lesson cards (localStorage)
- Every scene rebuilt against a fixed safe area (nothing off-canvas),
with animations that depict the narrated mechanism: contained
thermometer fills, a drawn dip-and-turn for the turning point, chaff
drifting off the swelling Maillard bean, crash/flick/stall on one
annotated RoR chart, per-brewer water/grounds anchored to each
silhouette's actual bowl, gauge needle sweeps instead of spinning
clocks; multiline labels replace out-of-frame text
- 9 new narration clips generated, 23 orphaned ones removed (31 total)
Co-Authored-By: Claude Fable 5 <[email protected]>
The admin nav-link reveal ran before the generated nav existed
(getElementById('nav-admin') was null only for admins — which is why
tests and non-admin checks missed it), aborting init before form wiring,
autosave, and loadPlans. initSideNav now runs first, the reveal is
null-safe, and wireCuppingLink degrades instead of crashing (its button
lost its home when After-the-Roast moved — it now lives at the end of
the Roast Log section with a pointer to the Roasts page).
Also: the Roasts page graph now honors the planner's °C/°F preference.
Co-Authored-By: Claude Fable 5 <[email protected]>
- The app header now wraps: brand + actions on the first row (actions
shrink and wrap instead of pushing Print off-screen), section pills
always on their own full-width row; the planner's sticky offsets use
a two-row --header-h via body:has(#plan-form)
- Roast detail gains a Plan selector (attach/detach any of your plans)
alongside the Machine selector; PUT /api/roasts/:id accepts
roastPlanId with ownership checks
Co-Authored-By: Claude Fable 5 <[email protected]>
One system: post-roast observations (weights, colour, DTR, cup notes,
'one change next batch', disproof) now live on the uploaded roast
(actual_roasts.after, migration 011), edited on the roast detail view
with weights/DTR prefilled from the .alog itself. The planner's screen
section is removed (the print worksheet keeps its hand-fill copy), and
the lot 'last refine' suggestion reads the note from both the new home
and legacy plans, newest wins.
Every roast now downloads two ways: the untouched original .alog, and
an updated supplemental copy with the app's data written back as valid
Artisan fields (weight, beans, roastingnotes incl. the LLM review,
cuppingnotes incl. linked cupping score/flavors) — serialized as a
Python literal so Artisan's ast.literal_eval re-imports it.
Co-Authored-By: Claude Fable 5 <[email protected]>
The section nav (Plan/Bean/Machine/Log/After) was an overflow-x:auto
strip with hidden scrollbars — undiscoverable on touch. It now wraps on
desktop and renders as a fixed five-column grid of equal pills at
≤720px (labels shortened to fit 320px), and the actions row lets the
autosave chip shrink/ellipsize so nothing can force horizontal overflow.
Co-Authored-By: Claude Fable 5 <[email protected]>
Machines:
- roasters table + actual_roasts.roaster_id (migrations 009): per-user
machines with a default; uploads attach to the default roaster and are
reassignable from the roast detail view
- Learning is now per machine: /api/roaster-profile scopes to a roaster
(default when unspecified) and applies the user's override tweaks
(milestone temps, TP time, pace factor) on top of learned medians
- /roasters page: machine list + analysis report showing learned vs
applied values with editable tweaks, typical RoR, and a plain-language
list of the adjustments applied to plans; planner honors a manual pace
override; evaluations use the roast's own machine profile
Mobile header redesign:
- The planner's header tools collapse behind a single 'Tools ▾'
disclosure (dropdown card) at ≤720px, keeping the fixed three-row
mobile header with no wrapped or scrolling button rows; desktop
renders inline via display:contents (same DOM, same handlers)
Profile pictures:
- users.avatar (migration 010) + PUT/GET/DELETE /api/account/avatar;
client-side centre-crop resize to 128px JPEG; avatar shows in the nav
chip on every page; included in full backups
Co-Authored-By: Claude Fable 5 <[email protected]>
- user_gear table (migration 008) + GET/PUT /api/gear (brewer keys
validated against the taxonomy, grinders deduped/trimmed)
- /gear page: silhouette multi-select for owned brewers (autosaves) and
a grinder list; nav gains Brewing → Equipment
- Brews page: with gear configured, the picker shows only owned brewers
(plus the edited brew's method) with a show-all toggle; grinder field
suggests your grinders; empty-gear state links to /gear
- Included in full backups, per-user export, and the OpenAPI spec
Co-Authored-By: Claude Fable 5 <[email protected]>
The Tools group (Plans, Ask the LLM, Prefill, Curves, Import/export,
Field IDs, Settings) acted on the open plan, not on navigation — it now
renders as compact action buttons in the planner header's existing
actions row (wrapping on narrow screens), and the left rail is pure
navigation, identical on every page.
Co-Authored-By: Claude Fable 5 <[email protected]>
The extraction schema gains cultivarGroup (typica/bourbon/ethiopian/
hybrid, classified by the LLM from coffee genetics); when the named
variety isn't in the cultivar table, deriveFields now falls back to the
group's first-crack anchor (number still comes only from the reference
table) instead of leaving 1.4 blank and breaking every downstream
calculation. deriveFields exported + unit-tested.
Co-Authored-By: Claude Fable 5 <[email protected]>
Free-text recipe (bloom, pours, timings) distinct from tasting notes and
next-time notes: new brews.recipe column (migration 007), form textarea,
shown under the recipe summary in brew history, carried by Brew again,
included in backups and the OpenAPI spec.
Co-Authored-By: Claude Fable 5 <[email protected]>
Brewing:
- roasted_beans + brews tables; /beans (bean management with LLM URL
prefill) and /brews (silhouette brewer picker across immersion/
percolation/espresso, recipe fields, auto ratio, 0-10 rating, tasting
notes); bean remaining weight derived from logged brew doses
- Green inventory lot form also prefills from a product URL
Navigation/UX:
- Side nav is now generated from one definition in nav.js, grouped
Roasting / Brewing / account, consistent on every page
LLM:
- 'Ask the LLM' chat drawer on the planner (stateless /api/plan-chat)
grounded in the plan, computed ledger, learned pace, and a new
roaster-behavior profile aggregated from uploaded .alogs
(/api/roaster-profile: TP lag, phase RoR, median milestone temps)
- The profile also feeds roast reviews and the planner curve's fallback
milestone temps
API platform:
- User-generated bearer tokens (rpt_…) with account-page management;
token requests skip CSRF; hand-authored OpenAPI 3 spec at
/api/openapi.json rendered by self-hosted Swagger UI at /api-docs
- Full-database backup export/import (admin) + per-user data export
Co-Authored-By: Claude Fable 5 <[email protected]>
- New server/llm.js consolidates ModelRuntime + model selection: admin
setting (app_settings.llm_model) > LLM_MODEL/PREFILL_MODEL env > first
available; used by both prefill and roast evaluation
- GET/PUT /api/admin/llm lists configured models and stores the choice
(validated against the list; empty = auto; audited); admin page gains
an LLM section with a model picker
- All user-facing 'Pi agent' wording is now 'LLM'; no_model error message
no longer references the pi CLI
- /roasts table: Review column removed (review lives in the detail view)
Co-Authored-By: Claude Fable 5 <[email protected]>
- POST /api/roasts stores the original .alog verbatim (multiple per plan),
parses it server-side, and kicks off an async zero-tool Pi agent review
(grade, highlights, concerns, next-batch suggestions, plan-vs-actual)
- /roasts page: table of historical actual roasts with review summaries;
row detail renders the BT curve as SVG with the plan curve overlaid,
shows the full review, and offers original-.alog backup download,
re-evaluate, and delete
- Planner's Reference curve drawer gains a multi-file finished-roast
uploader that attaches to the open (synced) plan
- Failed reviews record the error on the row and are retryable via
POST /api/roasts/:id/evaluate (e.g. once a model is configured)
Co-Authored-By: Claude Fable 5 <[email protected]>
New plan reset the form correctly but left the Plans drawer open over
it, so the reset was invisible until the user closed the drawer
themselves. Now closes it immediately, matching selectPlan().
Also adds a delete button per plan in the Plans drawer (server route
already existed) - deleting the currently open plan resets to a fresh
blank one, matching New plan.
Why-panel teaching callouts ("WHY CULTIVAR SETS THE CLOCK" etc.) now
default collapsed instead of wall-to-wall expanded on first load, with
an "Expand all" button in the header on both the planner and cupping
pages. Persistence flips from tracking collapsed-vs-default-open to
expanded-vs-default-closed to match.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Replaces the flat additive batch-size correction with a per-user
multiplicative pace factor learned from each account's own logged
roasts (shared/learn.js, GET /api/machine-profile), plus a lot-scoped
"last refine" auto-suggestion. Also fixes the reference data being
framed as "your own roasts" on a now-multi-user product, and
reclassifies the drying/Maillard sanity checks as informational since
they're algebraically derived from the DTR check rather than
independent (yellow = 0.56 x first crack, not entered separately).
Bug fixes found by an adversarial Opus review of the first pass:
- Unicode minus sign (U+2212) broke duration parsing against the
app's own generated refine-suggestion text
- Printed sanity-checks table still showed a bare pass/fail glyph for
the now-informational drying/Maillard rows
- Printed time-ledger box didn't show the pace multiplication step,
so it stopped reconciling by hand once pace != 1
- Field 6.4 (manual batch correction) was double-counted: excluded
from the learned-pace fit but added back after the multiplication
- Learned pace had no outlier rejection or hard clamp
- computeLedger had no test coverage
- Batch-size help copy overstated what the pace factor models (it's
a single blanket ratio, not conditioned on batch weight)
A follow-up Opus pass also caught the per-user profile cache
surviving logout/account-switch in a shared browser; fixed by
sweeping it alongside the existing plan-draft cleanup.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Highest severity: injected "?" buttons were hijacking the implicit
<label> association on 14 fields, so clicking the label text opened
the help dialog instead of focusing the input and screen readers lost
the input's accessible name. ensureLabelAssociation() now gives each
affected input an explicit id/for pairing before the button lands.
Also: milestones-ror/actuators topic placement is now text-matched
with a warning fallback instead of blind array indexing; every topic
warns if it resolves zero containers; dialog temperatures (milestone
table, RoR budget, worked example, both charge-setting tables) now
respect the °C/°F toggle instead of being stuck in °C; the print
worksheet's duplicate curve chart was missing the data-tempc
attributes its live counterpart had, so only its title/caption (not
its axis numbers) converted; the "?" trigger's touch target grew from
15px to a WCAG-conformant ~25px without changing the visual dot; two
genuine coverage gaps closed (a roast-log checklist topic, a
curve-reading topic); a worked feasibility example was restored; two
claims that didn't trace back to the source worksheet were corrected
(an invented "±15 seconds" sweet-spot tolerance, a wrong
moisture-to-time mapping); three cultivar cells regained detail lost
in the original transcription; a data-transcription footnote no
longer leaks into an unrelated topic; checklist items keep their list
semantics under VoiceOver.
Co-Authored-By: Fable <[email protected]>
Co-Authored-By: Opus <[email protected]>
Adds a help-dialog UI (field-help.js/field-help-content.js) surfacing
per-field guidance, expands cultivar and process reference data with
cup notes, sweet-spot width, and watch-fors, and adds a Makefile for
common npm tasks.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
The Machine Plan chart's axis ticks and first-crack band caption, the
print worksheet's axis title, and its "Your logged band" reference
column were all still hardcoded in °C. They're now data-driven
(data-tempc / data-band-lo / data-band-hi) and update alongside every
other temperature field when the unit is switched.
Plans are always stored in canonical °C; the unit is a client-side
display preference (localStorage) that converts Machine Plan planned
temps, Roast Log actual temps, and actuator expected BT on the fly,
including band-marker positioning. Toggle lives in Settings & sync.
Ports inventory management and SCA-style cupping scoring from
hope_roaster: lot tracking with audit-logged consumption, cupping
sessions with server-authoritative scoring and a live radar chart,
and links from the planner (draw-from-lot, open-cupping-session).
Also fixes the Blend/Single-origin toggle layout, replaces tooltips
with an in-context "why" teaching layer, and stages the planner UI
into pre-roast vs. post-roast phases.
snowspeederandClaude Fable 5