VideoState, Playlist, PlaylistItem, SavedView, ComfortPin, Tag and ChannelTag leave app/models.py
for app/reader/models.py; app.tables registers them. No schema change.
The nine Capture tables (files, jobs, folders, shares, links, quota, cookies, profiles, library
resume points) leave app/models.py for the unit's own module; app.tables registers every unit's
tables for the migrations, the test database and the ownership check. No schema change.
Backend
- A run's outcome write sits outside the job's own try: a database blip after a run that
succeeded no longer records it as failed (the row waits for the next start's sweep).
- An interrupted run keeps what it spent (the sweep sums its quota rows, as finish does).
- The Scheduler's ad-hoc row says running while any backfill runs, not only the newest.
- recent_runs reads each job's last runs off its index instead of ranking the whole table.
- An interval change is stored with its audit row in one transaction, then rescheduled.
- One API instance per database is the stated contract (.env.example, jobruns).
- Turning full history on for a channel already complete records no run.
- Migration 0109 round-trips interrupted and ad-hoc runs; the run words' DB CHECKs are
test-held to the enums; the account filter is one helper for the four activity reads.
- An audit row says how many of its request's log lines are still kept.
Frontend
- An "All" period reaches every row still kept (runs and admin actions outlive 90 days).
- Back/Forward is followed in place by the server log: a closing dialog no longer turns
Live off; a pivot still opens a fresh log.
- Runs moved over from the audit log read their one-liner in the reader's language.
- Filter choices or accounts that failed to load say so beside the filters, with a retry.
- An old server-log link keeps its account, text and range; a remembered retired page
opens its tab.
- A request whose lines are gone says so instead of linking to an empty list.
- The log lines and the history rows share one row frame, and the server log one list
frame with the other tabs; an empty list is announced; each day's quota is readable.
The audit log, the server log and the admin quota view were three pages (the quota one a Stats tab),
each with its own filters, and the background runs had no list at all. Now one page ("Napló" /
"Activity") reads all four, each still its own table on the server: a shared bar (text, a period of
1 h to 90 days defaulting to 7, an account) and each tab's own filters, the whole view in the URL.
The old page ids open their tab.
Server side: the audit log and the new run list (`/api/admin/runs`) filter and page like the server
log, through one keyset helper (`routes/_keyset`) the log now shares; an audit row keeps the request
it was made in (migration 0111), so its log lines are a pivot away, as a run's are (`lrun`). The
quota breakdown takes since/until and one account. The Scheduler rows get a history button to their
job's runs.
Also fixed on the way: re-saving the value in force on the Configuration page staged its override
without committing it (the commit sat inside the audit branch), so the save was lost.
An interval override was a row of its own table beside the one store every other admin knob uses; it
is now the `job_interval_<job>` sysconfig key (migration 0110 copies them over and drops the table).
The Scheduler page stays its one editor: the Configuration page and the setup wizard neither list nor
write those keys, since a write there would skip the reschedule.
Each ConfigSpec now states `applies` (required, no default). Every key is `live`, held by a scan that
finds no read at import time; a `restart` key is refused until the page can say so. A second scan
requires a reader for every key on the Configuration page — `sponsorblock_default` had none since it
was added, so it leaves the registry, the translations and the store.
What a job last did lived in the API process's memory and was gone at
every restart; the audit log kept a run only when it was manual, failed
or changed something. Now each run is a job_runs row written at its
start and its end in its own session: who started it, its outcome as
data, what it spent of the YouTube quota (quota_events.job_run_id) and
the lines it logged (log_events.run_id, filterable in the log console).
- a run a stopped process left open is marked interrupted at startup
- "run now" is a one-shot job on the scheduler, which now always starts
(interval jobs only with SCHEDULER_ENABLED); the manual actor is reset
after the run, since pool threads are reused
- the dashboard's last run comes from job_runs, so it survives a deploy
- runs leave the audit log: migration 0109 moves the old job_run rows
over (and back on downgrade); the audit page's run folding goes
- job_run_retention_days (default 30) prunes old runs in audit_gc
A /watch link may now open one of its owner's own folders (migration 0108: folder_id beside job_id,
exactly one set; a folder's links go with it). The page shows what is in the folder when it is
opened, newest first: a video or a song plays in the same player a file link has, above the list,
each keeping its own resume point; a document only downloads, never opens as a page. A file moved
out stops opening through the link at its next request. Password, expiry and stream-only work as on
a file link; the unfurl names the folder and counts its files, never their names. Source folders
and the Hub cannot be shared.
The folder menu gets "Copy link" and "Share with a link…" (the Share dialog's link half); the rail
marks a folder a live link opens.
Two fixes in the Share dialog on the way: revoking the last link no longer mints a new one at once,
and the new-link form goes back to downloadable after making a link.
A /watch link's token is the credential, and it sat in the database as written. Migration 0107
replaces the column with the token's SHA-256 (the lookup) and a Fernet-encrypted copy (so the owner
can copy the same link again); every existing address keeps working, and the downgrade decrypts
them back. Without a copy this key can read, a link still opens but its address cannot be shown:
"Copy link" then takes the newest link that has one or makes a new one, and the Share dialog says
so in that row instead of a broken URL. Making, changing and revoking a link are audited, never
with its token or password.
Settings gains an API tokens tab: a row per token (name and prefix; each
folder with its rights; last use and expiry, "expires soon" marked),
revoked from its menu behind a confirm. A new token is the Hub exchange in
one click, or folders and rights picked per folder (no Write where an
upload cannot land); expiry 30/90 days, 1 year or never (warned). The
token shows once in the same dialog, with Copy and a ready curl line.
The Hub folder closes the Sources in the Files rail, takes uploads and
moved-in files. A token about to run out warns its user once, a week
before (the Files GC's sixth pass); the notice opens the tokens tab.
A user can make a token for a machine of theirs (`/api/tokens`, signed-in only): `sfl_` + 32
random bytes + a checksum, only its SHA-256 kept, shown once. Sent as a Bearer header it is
honoured on six `/api/files` routes only — folders, list, file, hash, upload, delete — and there
only in the folders its grants name, with the rights they give (list / read / write / delete).
One check, `FilesCaller.may`, in every route; a folder or file out of reach answers like a
missing one; the files shared with the user are never in reach. A token beats a cookie on the
same request.
The Hub preset makes the user's one Hub folder (a new folder kind) and grants it every right.
Refused tokens are slowed down per client, a runaway machine per token; last use is stamped at
most once a minute; making and revoking a token is audited, and a token's uploads and deletes
name it in their audit rows.
- The queue keeps a done job by when it became done (ready_at), not by its last write: renaming,
keeping or moving an old file, or the GC's expiry warning, no longer brings it back as just finished.
- GC expiry and eviction re-read each file under the asset→job locks the user routes take, and commit
one file at a time: a delete, a re-download or a keep made while the GC runs is honoured, never a
second release that frees a file another user still holds.
- A file records the origin it was placed under (download_jobs.source_key, migration 0105): an edit
whose parent expired still goes back to its folder, not to Uploads.
- A folder a file is being put in is held (FOR SHARE) until it lands; a delete of it waits and takes
that file home too, instead of failing on the foreign key.
- The hash route reads at most two files at once (429 past that) and holds no connection meanwhile.
- 0104's backfill dates old files by their request date only: no notice prod wrote names its job.
- The old DOWNLOAD_ROOT still finds the files; a files root that is not a mounted volume stops the
worker and is logged by the API; the volume-copy recipe looks the project-prefixed names up.
- Stale comments about the /downloads mount and the Downloads page.
A download that failed for days and then went through sat days down the library, under its
request date, and read as missing. A job now records when it became done (`ready_at`), stamped
where the expiry already is (`folders.stamp_done`, `finalize_if_running`) and on an insert that is
already done; a check constraint holds every path to done to it. The library sorts and dates its
"Added" by it; the queue keeps the request order.
The migration dates a file from before by its "ready to watch" notice where the inbox still holds
one, else by its request date, where it already sat.
Files is the one library: every done file, the ones shared with me, or one
folder's, picked in a side rail (sources, own folders, the storage meter) and
mirrored to ?folder=. One table with every kind of value in its own column;
Copy link, Share and Save in sight, every action on a right click: play,
cut, rename in the row, move to an own folder or back home, keep, details,
re-download, delete.
The Downloads page keeps only the queue until the tray replaces it. The old
/api/downloads library routes are gone: list shared, upload, the details
PATCH, delete, the file and the usage now live under /api/files, and a
listed file says which source it came from. The job's "expired" flag is gone
with the state it described.
/api/files is the library's home: list a user's done files (by folder, searched, sorted, or the
ones shared with them), rename / move / keep / delete a file, make, rename and delete own folders
(a deleted folder's files go back to their source folders), and set how long any folder keeps a
file — done files shift by the difference, never closer than the GC's warning window.
An upload can name its folder (an own folder or Uploads), checked before the body is read and again
before the file lands; it is hashed while it streams, and any file can be hashed on first ask. The
download says the hash (Repr-Digest) once it is known; a new file at an asset's path forgets it.
Uploads, deletes, keeps, moves and folder changes are audited (ids only). The fileless "expired"
rows of the old retention are deleted by migration 0103: an expired file now leaves the library.
The shared route helpers moved to routes/files.py; the old /api/downloads library routes delegate
to the new ones until the client moves over.
The Files library starts here. A user's every download, clip and upload now
lands in a source folder made per origin (YouTube, Plex clips, uploads, one per
other site), decided once in downloads/folders.py at every place a job is made.
How long a file stays moves from the shared asset to the user's job: stamped
from its folder when it is done, never for a file the user keeps. The GC warns
each user once, then takes the expired file out of their library; the cache cap
never evicts a kept file, and an evicted one leaves every holder's library.
The media mount is /files now (FILES_ROOT, FILES_HOST_PATH, volume
siftlode_files); docs/self-hosting.md says how to carry an old volume over.
Migration 0102 places every existing job and carries each done file's expiry
over from its asset.
safe_abs_path takes a missing path and answers None, which drops one baselined
type error.
A suspension now blocks the user's downloads as quota.resolve already advised: the worker leaves
their queued jobs unclaimed until it is lifted (the confirm says so). A cut clip larger than the
remaining storage is refused with its own message (not "storage is full"), and a clip's quota
refusal reads as a clip's in both languages. The cancel poll compares to JobStatus.RUNNING.
- The cut file is measured against what its owner has left, for a YouTube range (whose fetch never
reads yt-dlp's size ceiling) and for a Plex clip (which fetches nothing), before the fetch and
again before it is placed; a refusal leaves the asset pending for a Retry.
- A silent Plex file is cut silent; a file the probe cannot read says so.
- A second job on a failed or renamed clip gets the same YouTube diagnosis and the fetched title.
- The "too long" refusal names the server's cap; a contract test holds the client's copies of the
clip limits to clip.py; the shipped player clients are pinned against the installed yt-dlp.
- One HDR transfer set and one rating_key lookup; job status words from JobStatus; the display-name
guard sees subscript writes and the clip route; stale comments and types follow the clip job.
An item can carry several credits blocks with mid- and post-credit scenes between
them, and Plex flags the one that really ends it. The player left at the first
block. Now the Plex-final block moves on (by the auto-skip or the Skip button), an
earlier block is only skipped to its end by the button, and with no final flag the
last block ends the item by the button alone.
Plex, library and /watch tell the operating system what plays (title, line, picture) and take its
play / pause / skip / seek through one hook (lib/mediaSession); the Plex player drives it with its own
seek and clock, so the lock screen shows the film's position, not the stream's. The YouTube player is
left alone: its iframe already publishes all of it, and the OS routes to the frame that plays.
The library and /watch players get the basic keys (lib/mediaKeys, the keymap's mediaPlayer rows) and
open where the viewer stopped (lib/resume): the library keeps the point on the server, per user and
library item (library_states, POST /api/downloads/{id}/progress, listed back on both listings), the
shared link in this browser. A started library row continues, its menu can start it over, and its
picture carries the cards' progress bar. The resume rule (too early / near the end = from the start)
is one function shared with the YouTube route.
The library dialog opens focused on its panel (Modal initialFocus "player"): on the ✕, Space closed
the film. While the media element itself has the focus, Space and the arrows stay its own controls'
— the browser acts on them whatever a listener does, and one Space played and paused.
- Every ffmpeg/ffprobe over a library file names the demuxers it may use
(edit.INPUT_GUARD; a downloaded thumbnail edit.THUMB_GUARD): a playlist-type
demuxer would read local paths named inside a user's upload.
- A re-encode ends on an even frame (edit.EVEN_FRAME): an odd 4:4:4/RGB source
failed its conversion; a clip stores the frame its file has.
- An upload counts against its owner's cap from the moment it lands (queued, or
kept after a failed conversion) — the footprint, the GC's cap and the admin
view read one ON_DISK predicate — and a conversion that outgrows the cap is
refused, keeping the original.
- The converted name keeps _a{id} (a dotted bare name once lost it and two
uploads shared a file); a converted file placed but not committed is removed,
and the orphan sweep enters .uploads; the LRU cap never evicts an upload.
- Ingest judges the extension the file is stored under; a failed conversion is
convert_failed / upload_missing, reports the "converting" phase, and notifies
as an upload. Job statuses in routes/downloads use JobStatus.
The chat cut removed the only WebSocket and CI went before it, but comments still leaned on both:
`?account=` was justified by the socket, job polling by an in-process WS registry, the e2e flows as
"proxy-safe (none use the WebSocket)", and pyrightconfig named a CI container the engine contract
test never checked. Each now says what holds today (the wallet gate, no push channel, the two
contract test halves, the image build's `npm ci`, the gate).
Without SMTP, registration marks the address verified with no proof, and an ADMIN_EMAILS /
ALLOWED_EMAILS address gets no invite row to gate it. So a stranger who registered such an address
(or a sibling spelling) before its owner kept their own password when the owner's first Google
sign-in adopted the row, activated it and made it admin, and could then sign in with it.
The adoption now treats a never-activated row like an unverified one: the password goes,
password_revoked_at is stamped and the session epoch bumped. An account already let in keeps its
password. The helper is renamed to say what it now judges (revoke_untrusted_preset_password).
test_google_callback.py drives the callback's account-matching branches through the real route,
none of which a test ran before: the squat (red on the old rule), adoption by mailbox, the
sub-collision and unverified-address refusals, the refused sign-in recorded in Google's locale, and
the link flow's conflict, mismatch and attach answers.
The code cited its plan instead of saying why: `E28 review J4`, `(R6 S3b)`, `S1 justified…`,
`deep-review round 5`. That plan lives in the private ops repo, so for every reader of this one the
citation was noise, and a sentence whose subject was the plan step ("round 2 showed…") hid its
lesson behind an id. Each one is gone and the "why" kept: a bare citation is deleted, a sentence
carrying a lesson is reworded around it ("a later attempt showed…").
Measured (`scripts/audit_metrics.py`, product code): 652 lines in 270 files carried an epic code by
the definition the next commit holds (507 / 241 by the old, two-digit-only one) -> 0. Across all
code, tests, scripts and configs: ~1100 epic-code lines plus ~320 sprint/round ids.
Behaviour unchanged, checked mechanically: every changed .py file has the same AST once docstrings
are blanked; css/sh/toml/yml/json files are identical once comments are stripped; the only
non-comment changes in TS/JS are test titles, one description string in a contract table, and five
ESLint messages, whose text is all the gates snapshot diff holds. Not touched: Markdown (docs and
the changelog keep their history), Alembic migrations, TV episode notation (`S2 · E5`), DESIGN.md
ids (`C1`, `B4`, `S3h`), tracker ids (`SEC-16`, `DX-17`).
One stale claim fixed on the way: test_hls_reaper's docstring said the player checkpoints "every
10 s whether or not it is playing", which plex/stream.py calls wrong; it now says what the code does.
The demo account — one shared user that whitelisted emails entered from
the login page, reset on a schedule — goes end to end (owner decision,
E34): demo.py and the reset job, users.is_demo and demo_whitelist
(migration 0098 deletes the demo user, its audit rows, the reset job's
override and notices, then drops the column and table; the downgrade
restores the empty schema), the /auth/demo entry, the admin whitelist and
reset, the Demo tab, the banner, the login page's "Try the demo", and
every is_demo branch in routes, userscope, scheduler and the UI.
require_human only ever refused the demo account, so its 87 uses become
current_user (owner-approved); require_write_scope keeps its own check.
The login page's "or" divider now shows only beside the Google button:
without the demo entry under it, an instance without Google would have
said "or" to nothing.
A contract test (removedFeatures) fails on the word anywhere in the
backend, frontend, e2e or scripts — seen red on a planted hit. Size
ceilings tightened for the modules that shrank; admin.py is back under
the limit.
E2EE direct messages, the chat dock, the Messages module, the message
WebSocket and its realtime registry, and the messages / message_keys
tables (migration 0097; the downgrade restores the empty schema). Prod
held two welcome notes, no message between people and one key — never
used (owner decision, E34). WebSocket now has no owner module: the lint
bans it everywhere. Privacy policy updated: it no longer promises E2EE
for a feature that is gone.
ruff.toml ignored E402 everywhere for main.py's logging setup. Measured: 28 hits, of which 11
had nothing to do with it — runner.py's logger line between its imports, models.py importing
canonical_email mid-file, a test constant above its imports, and main.py's own third-party and
mimetypes imports. Those are moved up. What stays after configure_logging in main.py is marked
per statement (noqa E402, logging is configured first), its comment now says why the order
matters (a line logged while importing, not the logger's creation). The register row goes, the
ruff key whitelist loses ignore, and a late import is red again. The T20 comment counts the
third deliberate print (scripts/e2e_seed.py).
The views, duration and title sorts had no index: every page read the whole videos heap and kept a
top-N (prod 350 ms a page, 550-780 for title). Migration 0096 adds one per direction and, for title,
per reader collation, in the exact ORDER BY feedquery builds. Localdev prod copy: plain sorts
180-310 ms -> ~1 ms, page 20 3-5 ms, narrow filters unchanged. view_count/duration are only written
by the enrich, which already rewrites indexed columns, so no HOT update is lost.
The title indexes are the first on an ICU collation, and postgres:16-alpine floats: collation.heal_drift
rebuilds them (CONCURRENTLY) at startup when Postgres reports the ICU version moved.
test_feed_sort_indexes holds every videos sort to an index and checks the rebuild.
GET /api/channels summed every subscribed channel's videos by reading each video's heap page
(681 MB on prod): 51-61 ms warm, but 2.25 s on a first open and 6.4 s under EXPLAIN once the
table had left memory. /feed/summary and /feed/years did the same (~140 ms warm, ~800 ms cold).
Migration 0095 replaces ix_videos_channel_pub_desc with the same key INCLUDING the columns
those queries read, so they are Index Only Scans over a 25 MB index. A test holds the queries
and the index's column list together.
The admin Configuration page did a db.get per setting (two for an unset one): it now reads
every override in one query (~40 ms of ~55 on prod).
A person's TMDB bio was fetched only when their page was first opened, off the request path, so the first
open showed no bio and the page never asked again (2 of 18,202 linked people were ever enriched).
- people_enrich scheduler job (every 10 min, 300 per run): linked people's bios from TMDB and their
portraits into the image cache, the most-credited people first. A failed TMDB lookup ends the batch.
- Portraits are resized through Plex's photo transcoder to a 320 box (~15 KB instead of ~170 KB; the
person-image proxy uses the same cache key, and a full-size original is removed as its small twin lands).
- The person response says `enriching` while a bio is on its way (a person the job hasn't reached yet).
- "Known for" is ranked by vote count, without guest visits as themselves (talk shows had crowded out the
films: Tom Cruise had 14 of 15), with directed works, from a 40-deep pool; the page shows up to 15.
Migration 0093 un-stamps enriched people so the job refetches them.
- A phased job result (backfill, people_enrich) now counts as a change for the audit trail; before, an
automatic backfill run that added videos was never audited.
- HU: the watch reconcile job reads "egyeztetés" (a leftover from the S5 sweep).
The /person fetch appends translations and keeps TMDB's Hungarian biography beside the instance-language
one (new plex_people.biography_hu, migration 0092). A Hungarian reader is served it, the English one when
TMDB has none. Already-enriched people are un-stamped, so their next open refetches them once.
The HU TV section lists every episode with its Hungarian title and summary, but the
overlay read only movies and shows, so every episode stayed English. It now overlays
the episodes too, and the HU tagline (new tagline_hu column, served on ?lang=hu).
Plex's HU agent returns a mixed genre vocabulary (Action beside Dráma), and a title
with no HU twin kept its English list, so the HU genre list showed Drama and Dráma.
plex/genres.py maps every genre word to one Hungarian word; the overlay writes it
into genres_hu for every row, and a HU genre filter is read through the same map,
so a saved filter still holding Action keeps matching.
Deep-review round 2, on round 1's own fix commit. It found MORE than round 1
did (28 raw, against 13), and most of it was mine -- a new piece of UI attracts
findings a backend-only diff never does. The pattern in almost all of them is
the same: a principle applied to the site that prompted it and not to its
siblings.
* `display_text` reached three of the FIVE writers of a user-supplied name.
`POST /api/downloads` and `POST /api/downloads/edit` still handed Postgres an
unbounded string, so a 300-character name pasted into the Add-download dialog
or the editor was a bare 500 (main.py re-raises every DataError but the NUL
case). Four of the seven reviewers found it independently. The bound now lives
in `models.py`, the three columns are DECLARED with it, and the guard for the
next writer is parsed rather than remembered: every `display_name` /
`display_uploader` assignment under `app/routes/` must go through the helper.
That guard immediately found two more -- the Google sign-in callback writes
the profile name into the same 255-wide column -- so it is six, not five.
* The connection release reached one of the THREE routes that read their own
body. The screenshot route (10 MiB, from a phone) and the client log sink both
still held a pooled connection idle-in-transaction for the length of the
upload, which is the invariant db.py sizes the pool on. All three now call
`db.release_read_connection`, whose precondition -- nothing pending on the
shared session -- is CHECKED rather than described, because a whole-session
rollback discarding a caller's staged write is a mistake this repo has already
made once. `test_body_limit.py` holds the rule for the next such route.
* The AST drift registry keyed its sites by (module, limit expression), so a
second route copied from the first collapsed onto the same entry and went
uncovered -- the exact hole the registry exists to close. Keyed per call site
now, and it carries the release requirement too.
* The dialog's own refusal was announced once and never again: two over-cap
picks leave an identical node with identical text, and an untouched live
region says nothing. It is keyed per pick now, the drop zone reads as invalid
and points at the sentence, and the reason is reachable from the control
rather than only from a one-shot announcement.
* Two tests rebuilt the blind spot they were written to remove: the dialog's cap
test sat nowhere near the boundary (a `>` loosened to `>=` survived it, which
is the mutation round 1 fixed on the server side), and the header-precedence
test only covered an exact-case duplicate while a differently-cased one would
have been COMBINED onto the wire rather than replaced. Both fixed, and the
merge now dedupes case-insensitively because header names are.
Also: the upload dialog's unstubbed usage query was firing a real request from
jsdom in a neighbouring suite and leaking a delayed "connection lost" toast
across test boundaries; `GET /api/downloads/usage` had no test at all though the
whole client-side check hangs off it; and the comment claiming the edge 413 is
readable now says what was measured -- readable to 200 MiB over loopback, a bare
`onerror` through a proxy -- so the pre-send check is named as the mechanism and
that branch as the backstop it is.
Every test added or changed here was mutation-checked. One mutation SURVIVED and
is recorded rather than papered over: the `tooBig` term inside `submit()` cannot
be reached, because `ui/Button` swallows the activation itself, so it is marked
as belt and braces instead of being described as the guard.
After Stop (or at the cap, or from Settings → Diagnosztika's past sessions)
one dialog drafts the issue: title, description, the steps drafted from the
session's markers and page changes (a marker without a label is named by its
time), the environment (version, browser, system, window, language). Beside
it, what goes with it: the browser and server logs — each reviewable line by
line, an unticked line is not sent — and the screenshots, removable, with
paste / drop / pick to add more.
- it saves as it goes (debug_sessions.draft, migration 0090): the rendered
markdown in the user's language, the excluded rows, and its own fields to
reopen as it was; a sent session is read-only
- the zip export (issue.md, client.log, server.log, screenshots) is built
from the saved draft by app/issues/bundle.py — the one definition the Gitea
delivery (S4) will share; .log, not .jsonl, which Gitea refuses
- the log line's level tags and clock move to components/logLineParts so the
review list draws a level the way both log viewers do
A session is one recording, capped at 60 minutes by the server. While it is
open, a request naming the caller's own session in X-Debug-Session is logged
at the session's level with every line tagged; a header naming anything else
is ignored. The browser's batches count a session as consent to its level and
keep the tag only on its own events. Screenshots are judged by their bytes,
capped in size and count, and served back to their owner only.
- debug_sessions / debug_screenshots tables, one open session per user
(partial unique index), log_events.debug_session_id indexed
- /api/debug-sessions: start (joins a running one), stop, list, get, events,
screenshots; /api/me reports the live session
- the hourly log job closes expired sessions and drops them with the logs
- logsink.serialize and routes._common.capped_body shared by their readers
YouTube's RSS goes dark per channel for hours (prod 2026-09-18/19: 393 of
396 feeds answered 404 for four hours, sticky per channel, so a retry does
not help), and nothing else read a followed channel's new uploads. A
channel whose feed has not answered for `rss_fallback_hours` (admin knob,
default 2, 0 = off) now gets one page of its uploads playlist through the
Data API (1 unit), at most once per window, yielding to the backfill
reserve. Spend is attributed to its own quota action.
Deep-review round 1 of E22.
- A NUL character (`GET /assets/%00` logs the decoded path) made Postgres
refuse the row and with it the sink's whole batch, every user's rows
included. `to_row` replaces NUL in every text column and in attrs.
- An erased account's lines kept arriving after purge_user's DELETE: the
erasing request's own access line, a queued row, a background task.
`logs.forget_user` runs first; later rows are stored without the account,
and the check and the insert share a lock with the erasure.
- An admin setting the whole process to TRACE logged every user's query
string. The query is now logged only under the user's own Teljes consent,
and decoded before redaction, so `%40` no longer hides an address.
- The browser log ingest has an hourly storage budget per account (5 MiB,
one notice when spent) and reads only a JSON Content-Type, so a
same-site text/plain form post cannot write into a user's log.
RateLimiter takes a per-attempt cost for the byte budget.
- The root level is applied under the elevation lock: two concurrent
changes could leave it at the base while an elevation was active.
- Tests for the running sink (level gate, drain on stop) and for the
admin log level applied at runtime and at boot.
A user can raise their diagnostics level to Részletes (DEBUG) or Teljes (TRACE) for 24 hours from a
new Settings tab, after a consent box that says what is recorded, that the server keeps it and for
how long (the instance's own retention), and that it lapses. Alap is the base and needs no consent.
- users.diagnostics_level/_until (migration 0087), written only by PUT /api/me/diagnostics, which
computes the expiry itself; the demo account cannot raise it. /api/me reports the level in force.
- The server logs an elevated user's requests at that level and stores every line of them
(logs.LevelGate on the stdout handler and the sink; the root is lowered only while an elevation
is active). Teljes adds the redacted query string to the request line.
- POST /api/diagnostics/events stores the browser's log as source=web rows: account from the
session, body capped while streaming, events/attrs capped, re-redacted, events under the
consented level dropped, timestamps clamped, 30 batches a minute, 409 once the level lapsed.
- lib/diagnostics.ts applies the level to the tab and streams only what happens after consent, in
batches, never its own request lines, with a keepalive flush on pagehide; stops on lapse (its
own timer), on withdrawal and on 409.
- Detail hooks at DEBUG: which control was clicked (test id, icon, app copy in English — never user
text, input values or a link's query), visibility/bfcache/resize, filter changes
(lib/filterLog.ts; typed search only at TRACE), player events (lib/mediaLog.ts; media errors are
warnings at every level). Offline/online at the base level.
- "Saját napló" viewer: live, newest first, level groups with counts, search, row details, copy
the filtered events as JSON lines, clear.
- Privacy policy and Terms exist in Hungarian and English (a contract test holds both languages to
the same sections); the policy now describes the server logs and diagnostic logging, self-service
deletion and password sign-in.
Fixed on the way:
- A sync auth dependency ran logs.set_user on a copied context, so the account never reached a
handler's own log lines (measured); a per-request mutable scope carries it now.
- The Settings tab rail left a phone no room (clipped labels, buttons over text): below `sm` it is
a scrolling row above the content.
- PUT /api/me/preferences took a bare dict body; it is a Pydantic model now.
The log now has a stored, queryable copy for the diagnostics console:
every WARNING and above plus the request line of every request, kept
for log_retention_days (default 7, no "forever") and pruned hourly by
the new log_gc job.
Rows are built on the logging thread from the same redacted fields as
the stdout line, queued without blocking and written in batches by one
background thread on its own connection. A full queue or a failed write
drops and reports instead of retrying. Flood control keeps 30 of any
one message per minute plus a count of the rest; anonymous request
lines are capped at 60 a minute per client, and the client address is
never stored. Deleting an account deletes its stored events.
Measured on prod first: a YouTube RSS outage wrote 11k warnings in four
hours, one per channel per poll. The poll now writes one warning with
per-reason counts and a few examples, and a failure on our side gets
one traceback per poll instead of one per channel.
users.last_login_at (migration 0085, no backfill) is stamped by _establish_session,
which the Google callback now shares with password login instead of repeating its
session setup inline. The demo entry does not stamp its shared account. The admin
user list carries the field.
A failed Facebook download read "Download failed": yt-dlp words the login wall
(18+ reel, closed group) as "Cannot parse data", which no category matched.
Measured on dev with the owner's cookie before writing a rule — the same text is
the login wall without a cookie and a refused cookie with one.
- errors.diagnose(msg, url, cookie_attached) tries a host-scoped catalog first
(all-of markers, cookie state, link fragment) and falls back to the category.
First entries: facebook.login_wall, facebook.cookie_rejected,
facebook.group_post_video. The module doc is the recipe for the next entry.
- download_jobs.error_detail (0084) + backfill of the stored Facebook parse
failures that were unknown.
- The pipeline's failure branch passes the link and whether the cookie went
with it; resume and redownload clear code and detail (redownload never
cleared error_code).
- A job meeting an already-failed asset got the text but no code, so it read
"Download failed" where the first job said why. It is diagnosed again, with
the cookie state from credentials.is_authed_asset (now shared with the
cookie gate); the redacted network text classifies back to network.
- FE: failCopyKeys gives the row and the failure window one lookup — catalog
copy, else the category. Copy in EN and HU; a contract test pins every
backend entry to both locales.
Deep-review round 3 of bug hunt 2026-09.
- The YouTube and Plex progress bodies are bounded by POSITION_MAX_S (the int4 position columns): a
finite position of 2^31 or more was a DataError, i.e. a 500 — the sibling of round 2's non-finite
fix.
- plex/stream.py promised "a periodic reaper", but reap_idle ran only inside the NEXT session
request: after the last playback an abandoned session (its ffmpeg still working, its directory a
whole segmented copy of the film) lived until someone pressed play again. A lifespan task reaps
every 60 s.
- Tests for what round 2 left untested: the 422's body keeps the list shape the UI localizes, and
stream_session refuses non-finite start/aoff.
E24 S1. The anti-paralysis answer layer, served from our own Postgres with no
upstream call: one hero suggestion, one CROSS-SOURCE resume rail, and the user's
own comfort shelf.
Survey first, because the roadmap line was wrong about the starting point: Plex
already ships the whole trio (E5's Discover rails, including a reshuffling
Surprise me), the YouTube side ships none of it, and nothing anywhere answers
the question across both. The data already existed on both sides — VideoState
and PlexState are structurally identical, the feed has an `in_progress` filter
and a seeded shuffle, and the YT player already resumes position.
- app/discover/budget.py — the time bands. "How much time do you have"
(10 perc / 30 perc / egy este / mindegy) replaces 15/D's mood questionnaire:
deterministic, answerable from a column we have, and it picks the source for
free. Matched against REMAINING time, so a film with 8 minutes left is a
ten-minute answer. An unknown runtime fits only the unbounded band.
- app/discover/cards.py — a thin unified card that carries the NATIVE feed /
Plex payload, so the page renders with the existing card components instead of
a third one. `aspect` is the one new field: a rail needs to know poster from
16:9, and that is a property of the artwork, not of the module.
- app/discover/queries.py — the SQL, reusing feedquery and plex.rollup rather
than the Plex Discover ROUTE, whose LibraryFilters narrowing belongs to a
surface that has a sidebar. One seed drives both halves so a reroll moves
them together.
- app/discover/hero.py — resume -> comfort -> surprise, with the chosen tier in
the payload so the client labels it honestly instead of guessing.
- comfort_pins (0083): five nullable CASCADE-ing FKs under num_nonnulls(...) = 1
rather than a flag or a polymorphic string ref. A flag cannot order a shelf or
hold a channel; a string ref would let a deleted library item leave a dead tile
on the one surface whose whole job is to be reliable.
Two honest-state rules the tests pin: the surprise draw excludes anything
already started (the resume tier may have skipped it for not fitting, and
offering it back as "Meglepetés" would contradict the question), and which
source leads the draw follows the seed, so the Plex library can actually answer
"egy este".
Gate green: ruff, ruff format, 3x tsc, eslint, prettier, knip, vitest 770,
pytest 987.
Twelve mails carried hardcoded English while the app itself has enforced HU/EN parity since E5.
A mail is the one user-facing surface nobody can re-render by switching the UI language: it arrives
once, in whatever language we picked, and stays that way. That made the epic's own headline mail —
"one mail, one click" — English for a Hungarian reader.
The text moved to a catalog (`mailtext.py`) with a `text(key, lang, **vars)` renderer, and the parity
rule is now a test rather than a habit: every key must carry both languages AND the same
`{placeholders}`. A translation that drops `{link}` sends a mail with no link in it; one that invents
a placeholder raises inside a background task nobody is watching. Both are silent in production.
Every `send_*` takes the language as its second argument, resolved by the REQUEST handler and never
inside the mail — these all run as background tasks, after the request's session is gone, which is
the same reason the approval mail computes its link in the handler.
Where the language comes from, in order: the account's own preference; failing that, the language the
person asked us in; failing that, English. The middle one is new — `invites.lang`, migration `0082` —
because the mails an access request produces go to an address that usually has no account yet, and
the signal does exist: the page their form was submitted from, or the locale Google reported for a
sign-in we refused. Captured rather than derived, for the third time in this epic (`google_email` in
S3, `preferences.language` in S6 taught the same lesson): a signal that exists only in one moment
cannot be backfilled later. So the admin is asked in exactly one place — the invitation to an address
they typed in themselves, where nobody has ever told us anything.
That send is now a dialog rather than a bare click, which is also where the "you already wrote to
them" warning belongs — it used to live only in a tooltip the admin may never hover, guarding the one
action whose failure mode is mailing a real person twice.
S6's registration `lang` field turned out to be dead on arrival: the backend accepted it, the
frontend never sent it, so every account has been storing nothing. The request builders now stamp the
UI language themselves — a signal that has to be passed in by every caller is one every caller can
forget, which is exactly how it shipped unset.
Admins are notified one mail per language group instead of one blast to everyone. They are people
with their own UI language; a single message would have to pick one of them and be wrong for the rest.
Also removes a duplicated `invited_at` column declaration on `Invite` (two identical definitions
landed in S6).
The funnel's common ordering made a person open TWO messages: one saying their access was approved,
another asking them to confirm their address. The approval mail now branches on what they actually
still have to do — carry the confirmation link, point at sign-in, or say a confirmation mail follows
once they register — so the usual case is one mail and one click. A mail that guesses wrong is worse
than a plain one: telling someone to sign in while their address is unconfirmed contradicts the
screen that will refuse them.
That embedded link lives 7 days rather than the usual 24 hours. Nobody sits waiting for an approval —
it can easily be read on Monday after a weekend, and an expired link spends the one-mail promise and
puts them back on two. The risk is small: the token only ever confirms an address; it signs nobody in
and sets no password.
The most likely root cause of the 2026-09-03 report that a confirmation mail "never arrived" is here
too, and it was never a delivery failure: re-registering an address that already had an unconfirmed
account returned in SILENCE — same cheerful "check your inbox", no mail sent at all. It now re-sends
the link. Nothing new leaks: the response and its timing are unchanged, and the public verify/resend
already does exactly this for the same input.
Whitelisting stays silent, and that is a decision, not an omission. The operator opens a door as a
possibility — usually already agreed with the person — and a row can sit approved for weeks with
nobody registering; a mail fired automatically there would be nudging someone who never asked.
Telling them is its own act on its own button, `invited_at` records it so a second click is a
decision rather than an accident, and the button only appears where nobody is behind the address yet:
"the door is open, come in" is meaningless to someone already inside.
Two things UAT caught that no test would have. Every fresh registration appeared in BOTH admin lists
at once, because the stuck-sign-ups filter keyed on "unconfirmed" — the rule is now "the address may
sign in, and this account still can't", which leaves an undecided request with the queue that owns
it. And the invitation button was offered on every approved row, half of which are existing users.
Also: the confirmation mail finally says what its link's rules are (once, 24 hours) — the reset mail
always has, while this one left the reader to discover both by meeting a dead link. The card's
"didn't get the email?" moves up under the form, and the three waiting messages name the spam folder,
which is where a first message to a fresh recipient sometimes lands.
Registration now records the language the sign-up page was in. Nothing reads it yet — the bilingual
templates are S7 — but it is the only moment that choice exists anywhere the server can see, and a
preference nobody recorded cannot be backfilled later.
UAT on localdev: re-registering re-sent the link (one token became two, one account); approval issued
a 7-day token and moved the account into the stuck list with only a re-send; whitelisting mailed
nobody; the invitation recorded itself and the row then warned that sending again would write twice.
Test data removed.
An account can now hold two true addresses: the one its owner typed, which a Google
sign-in no longer overwrites, and the one Google answers with. Only the first was
stored, so Settings could say 'Connected' without saying connected AS WHOM -- exactly
what someone who registered with a +tag needs to know.
users.google_email (migration 0080) records Google's answer at sign-in and at an
in-app link. /api/me exposes it ONLY when it differs from the account address, so the
comparison lives in one place and the majority, whose two addresses are the same
string, see no change at all. Rendered in the account menu as a muted third line under
the identity, and in the Settings sign-in row under the existing hint.
A short 'Google' label rather than a brand glyph: the app draws one icon set (lucide),
and a generic mail icon next to an email address says nothing. It also reads to a
screen reader, which an unlabelled logo would not.
No backfill is possible -- Google's answer is only known at sign-in -- so accounts
linked earlier stay NULL until their owner next signs in with Google.
Gmail delivers x.y@, xy@ and xy+tag@ to one inbox; the invite gate only lowercased.
On 2026-09-03 that turned an approved user's Google sign-in into a silent SECOND
access request -- OAuth returns the account's stored primary, which need not be the
spelling they typed -- and left prod with two user rows for one mailbox (ids 13, 14).
The +tag case fails every time: a Google account cannot contain +, so OAuth can only
ever answer with the base address.
Mail still goes to the address AS TYPED; only identity matching uses the canonical
form. utils.canonical_email is the single definition, and a model validator derives
users.email_canonical / invites.email_canonical on every write, so no call site can
forget it. Scope is narrow on purpose: dots are significant on most providers and the
Gmail rule does not hold for Workspace custom domains.
Swept every sibling of the same principle, not just the reported one: is_allowed,
upsert_pending_invite, the Google-callback adoption, _register_account (so no new
duplicate pair can be created), _activate_user_by_email, admin _accounts_for (an
invite backing a live user under another spelling used to look deletable) and
add_invite, plus purge_user, which now erases every canonically-equal invite instead
of leaving one behind that still grants access.
Migration 0079 backfills both columns and indexes them NON-uniquely: prod already
carries canonical duplicates (approved invite pairs since 2026-06-16), and merging
accounts is a judgement call about people's data, not something a migration does on
its own. A test pins the backfill SQL against the Python helper.
SEC-10 revoked any password set while an account's email was unverified, on both
the Google-adoption and the admin-approval path. On the approval path that was
redundant: password_login refuses an unverified account before it ever reads
is_active, so approval hands a pre-emptor's password nothing it could use. All the
revoke achieved there was destroying the password a legitimate registrant had just
chosen -- and since the admin normally decides before the user reads their mail,
'your password was cleared for security' was the DEFAULT experience, not an edge
case. The guard stays on the Google-adoption path, where the real owner does attest
ownership.
password_reset_required and the ok_reset banner stay: accounts stripped on prod
before this change have no other way back in.
Two existing tests pinned the old rule, so they are rewritten rather than deleted,
and the SEC-10 recovery pack now builds the revoked state through the surviving
producer.
Two independent bugs made an email+password registration unrecoverable in prod.
The verification link did nothing. urlWithParams/clearParams rebuilt the URL as
pathname+query and dropped the fragment — but the one-time auth tokens ride the
fragment (#verify=/#reset=, SB3), and the providers that mirror nav params
(useUrlMirror in FeedFilters/Playlists/Plex/Channels) sit ABOVE the lazily-loaded
<App> in main.tsx. Their mount effect replaceState'd the token away before the auth
card's chunk had loaded, so the card read an empty hash: the link opened a bare
sign-in page, no request, no banner. Both helpers now carry the fragment over.
The account was then locked out for good. SEC-10 nulls password_hash when an
unverified account is attested (here: admin approval landing before the user could
verify — which the bug above made impossible). That is indistinguishable from a
Google-only account, and reset eligibility keyed off password_hash alone, so no
reset link was ever issued: no password to sign in with, and no way to get one.
users.password_revoked_at records the revocation, reset eligibility honours it, and
setting a password again (reset-confirm or set-password) clears it. An account that
never had a password stays ineligible on purpose — its owner adds one from Settings.
/auth/verify now also answers password_reset_required so the card can say why the
registered password stopped working, instead of leaving the user at the uniform
"Invalid email or password". Telling the holder of a valid verify token leaks
nothing: they have already proved they own the mailbox.
The 0078 migration backfills the stamp for rows carrying the exact lockout signature
(local, non-demo, no password, unverified), so existing victims recover unaided.
Adds a movie-scoped "Playback" filter to the Plex feed so a user can isolate items by how
they play — the groundwork for collecting HW-transcodable (or no-transcode) versions.
Four classes derived from the stored `playable` + a new `transcode_hw` facet:
- Direct Play (direct), Direct Stream (remux)
- Transcode · HW — playable=transcode AND the iGPU can hardware-transcode it
- Transcode · SW — playable=transcode, software libx264 only
The HW/SW split is precise, not codec-only: a 10-bit/HDR HEVC is HW-decodable yet breaks the
VAAPI nv12 pipeline (the same case plex/stream special-cases), so it must count as SW.
Backend:
- New PlexItem columns (migration 0077): transcode_hw, video_bit_depth, media_probed_at.
Migration backfills transcode_hw codec-only so the filter works immediately.
- `plex.sync.transcode_hw_capable` is the single classifier; the HW-decodable codec set moves
to sync as `HW_DECODABLE_CODECS` and plex.stream imports it, so the catalog facet and the
runtime encoder choice can't disagree by codec.
- Bit-depth backfill (`_probe_media_batch`, wired into sync): probes only HW-decodable transcode
items' video Stream for bit-depth + HDR and refines a 10-bit/HDR HEVC to software. Bounded
per run, watermarked by media_probed_at. A codec change resets the watermark (stale probe).
- Filter + self-excluded facet on a shared `playback_class_expr` CASE (movie-leg only — a show
is an aggregate with no single class). transcode_hw threaded onto the feed/detail/card JSON.
Frontend:
- Multi-select "Playback" Combobox in the sidebar (movie scope, facet-empty drop-out), active
chips, 4-way corner-dot badge (green/blue/amber/red) + tooltip, and an info-page playback chip
with a codec · bit-depth subtitle. New `bg-info` (blue) token, i18n HU/EN parity.
- Fix a regression where the card's playback dot lost its hover tooltip: the full-card reveal
overlay swallowed the cursor; the dot now sits above it (z-20).
Gate: backend full lane + frontend typecheck/lint/knip/prettier/vitest all green. Browser-UAT'd
on localdev (filter narrows to the SW set, facet counts, badge colour + tooltip, info chip).
peter