The proxy sign-in believed Cf-Access-Authenticated-User-Email from any
address in trusted_proxies. On Tower that address is the Docker gateway,
so any container there could name itself anyone (docs/sso.md said as
much, and CLAUDE.md listed it as a known gap).
With [web] access_team and access_aud set, a proxied request must also
carry a Cf-Access-Jwt-Assertion that verifies against Cloudflare's keys
(RS256 only, this application's audience, the team's issuer, not
expired), and the name comes from its email claim. The keys are fetched
at start and again when a token names an unseen key, at most once a
minute, so made-up key ids cannot make every request a request to
Cloudflare. While the keys cannot be had, proxied sign-in is refused;
password and token sign-in are unaffected. Both settings empty, nothing
changes.
jsonwebtoken does the checking, on the aws-lc-rs backend already in the
tree through rustls. Tests sign with throwaway keys in tests/data: a
valid token, another app's audience, expired, a forged signature, HS256,
alg none, the refetch limit, and keys that cannot be fetched. Checked
live on a scratch daemon: the header alone and a forged token got 401,
the admin token still signed in.
vouched_name takes the peer and headers rather than the request: a
&Request held across the new await made the auth middleware's future
unsendable, as a body is not Sync.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The daemon printed http://<bind>/?token=<token> at every start. The token
signs in as the admin, and in the container that line lands in docker
logs, readable by anyone with Docker access on Tower. It now says where
the token is kept instead; config.toml already has it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The scan's events reach everyone, so every browser showed "<feed>: N new" and
"Scanning…" toasts, and refreshed, for everyone's feeds. Now a feed's row, and
its folder's, carries a spinner between feed_start and its done, skip or error;
the list refreshes only for the reader's own feeds; the scan toasts are gone, and
"Downloaded" is said only for a file on screen.
"Check every feed" from the web UI sent a scan of every feed on the server.
Command::Fetch takes an optional `feeds` list -- those feeds and the feeds
inside any OPML among them -- and the web fills it with the asker's
subscriptions. The schedule and the CLI send none, meaning every feed.
Closes#37.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Phase 3 of #18. Two tables: catalogue (each feed's config::Feed as JSON, so a
new feed setting needs no column) and settings (general: the five server
settings the admin page edits). config.toml keeps what is needed before the
database is reached, or decides who gets in: paths, [torrent], [web].
ipx still runs from one in-memory Config, assembled at start from both
(assemble_config). The eight places that saved config.toml and re-read it now
call Ctx::store_cfg, which writes the database and swaps the copy in memory; the
first-run web token, which is config.toml's, is written there.
The first start on a database with no catalogue imports config.toml's feeds and
settings in one transaction whose first insert is the settings row, so two ipx
starting at once cannot both import; it then trims config.toml, keeping the
original as config.toml.pre-database. After that, feeds written into the file are
ignored with a warning. copy-db skips it, and copies both tables.
Rehearsed on a clone of production's database with production's config: all 130
feeds imported, the file trimmed, and the feed list, settings and directory
identical to the live server's.
Postgres connections now ask for no notices. Every CREATE ... IF NOT EXISTS on an
existing table sends one, eleven per open; sqlx logs them, and
tracing-subscriber 0.3.23's per-layer filters then dropped the next line ipx
logged -- the import's own message went missing that way. Proved by toggling it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
- IPX_DATABASE_URL (postgres://...) picks the database; unset, it is the SQLite
file as before. Passwords are taken out of anything logged.
- `ipx copy-db <state.db>` copies every table into the empty database the URL
names, in one transaction, and moves the id counters past the copied ids. A
copy of production went across in 14s with every count and column
fingerprint identical.
- With IPX_TEST_DATABASE_URL set, each test gets a Postgres schema of its own;
all 79 pass on both databases. Fixtures write booleans as true/false.
- Sorts say where an item with no value goes (NULLS FIRST going up, LAST going
down): SQLite counts NULL as smallest, Postgres as largest, so "largest first"
on Postgres led with every item that has no file. Tested on both.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The last nineteen functions move to SeaORM: recording feeds, items and
enclosures, managed OPML feeds, folding WordPress's repeated files, and handing a
Patreon creator's files to its shows. Two SQLite-only forms go: GLOB becomes a
LIKE with the underscore escaped (broader, harmlessly: the fold still keys on
`_=` and digits), and UPDATE OR IGNORE becomes an UPDATE ... WHERE NOT EXISTS.
The two transactions are SeaORM transactions.
With nothing left on it, rusqlite goes, with the SQL schema and migrate(). The
entities are the schema: create_missing makes whatever tables and indexes a
database lacks, from them, with CREATE ... IF NOT EXISTS. Production's schema
already has every column migrate() added and none it dropped.
Not SeaORM's schema sync, used until now: despite its docs it drops a unique
index the entities do not describe, so it dropped users_name_lower on every open.
Every `ipx` command then took a write lock, and against a daemon busy writing,
`ipx status` -- the healthcheck -- failed 7 times in 15 where the old code
failed none. Now 15 in 15, as before. On Postgres it would not have started.
WAL is set only when a file is not already in it: setting it takes a lock that
cannot wait out a busy daemon.
Checked on copies of production: a forced scan of all 162 feeds against the real
feeds with no database errors; the feed list, filters, sorts, search and the
reaper's candidates against the old code on the same data, earlier in the
branch. The column comments from the SQL schema move to the entities.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Twelve enclosure functions move to SeaORM: recording, the download queue,
marking done or failed, requeueing, and what the reaper may delete. INSERT OR
IGNORE becomes ON CONFLICT DO NOTHING; the reaper's read verdict is true or
false rather than 1 or 0, which Postgres would type as a 32-bit integer and
refuse to read as an i64; `read = 1` and `flagged = 1` test the booleans
themselves. retention::run and its callers (reap, rm, retire_group,
retire_stranded) become async.
The reaper deletes files, so it was checked on a copy of production against the
old SQL on the same file: all 2,195 candidates, identical and in the same order.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The item list, its counts, filters, sorts and search, positions, pins and
mark-all-read move to SeaORM, as SQL written for both databases:
- Parameters are gathered as the SQL is written (Args), so only what a
statement uses is bound. rusqlite needed every one mentioned, hence the old
`?1 IS NULL` and `?2 = ''`; Postgres refuses a parameter it cannot type.
- Yes/no columns are tested as booleans (NOT coalesce(s.read, false)) and
written as true, not 1; SQLite reads true and false as 1 and 0.
- The last tiebreak of the sort is the guid, not SQLite's rowid, which Postgres
lacks. Only items with the same date change places.
- set_position names entry_state.duration beside excluded.duration.
- The status callback on the control socket returns a future, as the counts
are now a query.
Checked on a copy of production against the live server: 42 of 48 lists
identical; the other six differ only in how ties fall, or because the test
daemon cleared paths to files this machine does not have. Run on the same file,
every filter's count matches the old SQL exactly.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Twelve subscription functions move to SeaORM. Lookups use the entity API; the
joins, counts and upserts are SQL written to run on both databases: $n
parameters, ON CONFLICT DO NOTHING in place of INSERT OR IGNORE, and
CASE WHEN on the yes/no column itself rather than comparing it to 1, which
Postgres would refuse for a boolean. INSERT ... SELECT ... ON CONFLICT gets a
WHERE true, which SQLite needs to tell the two apart.
Checked with a daemon on a copy of production: the feed list, read through the
new code, comes back with every feed and its settings.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The fourteen user and session functions move from rusqlite to SeaORM and become
async; their callers await them (auth, admin_user, user_cmd, the account
handlers). Checked against a copy of production, where the yes/no columns are
still INTEGER: the admin flag reads back right.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The first step of moving to SeaORM (#18, phase 1). Nothing a user sees changes.
- src/entity.rs: the seven tables as SeaORM entities, matching the SQLite schema.
Strings are Text, as the columns are; yes/no columns are bool, which is BOOLEAN
on Postgres and stays INTEGER in the existing SQLite file (sync notes the
difference and leaves it alone).
- Db holds a SeaORM connection to the same SQLite file beside the rusqlite one;
functions move to it one at a time, and rusqlite goes with the last of them.
- db::sync creates what a database is missing from the entities (SeaORM's
schema-sync, experimental, so sea-orm is pinned to ~2.0), plus the two indexes
an entity cannot express. Checked against a copy of production: it added the
lower(name) index and changed nothing else.
- Test databases are now built from the entities alone, in a temporary file
(two connections to one ":memory:" are two databases), so every test also
checks that the entities describe what the queries need. That caught the one
difference: finding a user by name relied on COLLATE NOCASE, which Postgres
lacks; it now compares lower() on both sides.
- rusqlite steps back to 0.39: 0.40's libsqlite3-sys is newer than sqlx accepts,
and only one may link SQLite. It goes away at the end of this phase.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Inter (#11): the pages are set in Inter's variable font, served from the
binary at /inter.woff2 as the icon is, with its OFL licence beside it in
web/. Classic keeps Lucida Grande, the 2004 app's face.
Double audio (#12): WordPress numbers each audio player on a page by
adding ?_=N to its file's URL, so a post that embeds the file it encloses
listed it twice, and it was downloaded twice. The parser keeps the first
of an item's enclosures that differ only by that number. At startup the
repeats already stored fold into the first; where only the repeat had
been downloaded its file moves to the first rather than being deleted.
Pin heading (#13): the rows' icon buttons kept the browser's side
padding, which pushed their 16px icon 3px right of centre, and the
heading's icon sat at the left of its column. Both are centred now, and
the heading row takes the pixel of border the rows have, so every
heading sits over its column.
Closes#11, closes#12, closes#13.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Almost no blog names a category the Directory can use, so a feed can
carry one of its own in config.toml, set by an admin in the feed's
settings and used when the feed names none. The feed's own iTunes
category still wins. The field offers the categories the Directory
already shows, so a blog about games joins Games rather than starting a
second chip. Setting it on a feed from an OPML promotes it to config, as
any other shared setting does.
Closes#10.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
davewiner's OPML left config.toml before retire_group existed, so its
derived rows were skipped by every scan but never cleared. At startup the
daemon now retires every group whose parent is gone from config. And
retire_group unmanages a feed that has its own config entry instead of
dropping it: eleven of davewiner's were promoted without being unmanaged,
and dropping them as derived would have deleted their entries. That also
covers removing an OPML or Patreon subscription from the page.
Closes#3.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Feeds take their channel's first <itunes:category> into a new feeds.category
column; the migration drops ETag and Last-Modified once so every feed re-reads
on its normal schedule and picks one up. /api/popular and /api/directory carry
category and podcast (any audio or video enclosure). Directory becomes a grid of
cover-art tiles under a chip rail: All, Podcasts, Blogs, and a podcast's
categories once Podcasts is picked. Popular and Add a feed keep their rows.
Closes#4, closes#5, closes#6, closes#7.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
- Stop scanning an OPML/Patreon feed's derived rows once nobody subscribes
to it; retire them (drop or orphan) the way sync_group already does when
the list itself drops one. This is what let 922 defunct davewiner feeds
keep scanning hourly after the OPML left config.
- Repair feed XML with a bare `&`, and give a plain reason (moved web page
with its new address when linked, or nothing yet for an empty body)
instead of a raw parser error.
- Show a failing feed's plain-English reason and next step (Unsubscribe /
Use the new address) in the sidebar and on its own page, once it has
been down a day.
- Fix four small UI bugs: show-note links open in a new tab, video files
play as video, an opened item no longer disappears from the Unread tab,
and Subscribe/Unsubscribe get their own icons.
- Fix Settings disappearing for non-admin accounts: it was hiding the
whole modal instead of just the admin-only parts (Users, the editable
schedule/quota, Save), which are the only parts the server actually
refuses them.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DmQfE1eFPApnXWyPHBWqUA
status is a terminal event. Broadcast, the healthcheck's answer ended any
ipx fetch that was watching a scan, which stopped reading at the next probe
while the scan carried on. It could not happen while status waited behind
the scan; answering it at once made it happen every 30 seconds. Each
connection's writer now takes private replies beside the broadcast, and
the test checks another client hears nothing.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TAC7sLVqfKmY6rsTLXzNgk
The worker runs one job at a time, and status was one of its jobs, so the
Docker healthcheck waited behind the startup scan (54 seconds of it after
the last deploy) and timed out at 5. Any scan or download longer than
three probes would have had a working daemon marked unhealthy. The socket
now answers status straight away; everything else still queues. A test
fills the queue and checks status comes back anyway.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TAC7sLVqfKmY6rsTLXzNgk
An account made by hand before the proxy was set up is called what it was
given ('rays'), while Cloudflare Access vouches for an email address. With
auto_create_users on, the first visit through the tunnel would make a
second, empty account. Renaming keeps the id, so feeds, read state and
admin rights go with it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TAC7sLVqfKmY6rsTLXzNgk
users.created comes back, beside a new last_login, for whoever maintains
the server. A password sign-in, the token link and a request through the
proxy all count, recorded to the hour so the proxy's per-request vouching
is not a write each time. Settings -> Users and ipx user list show both.
The three user queries now share one row mapping.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TAC7sLVqfKmY6rsTLXzNgk
Works through TODO.md from the 2026-09-12 over-engineering audit. Drops the
entries.read/flagged/position columns (migrate() removes them from older
databases), migrate_opml_children, the legacy interval_mins key, the
contrib/ systemd units, test-only Db wrappers, a duplicate token generator,
redundant logbuf visitors, unused page state and CSS, and the infer, dirs
and tokio-stream dependencies. The icon is served once as /icon.png instead
of inlined four times, taking about 94 KB off the two pages.
The adoption's subscription half was not dead: it gives a fresh install's
first admin the config's feeds. It stays as adopt_catalogue, now tested.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TAC7sLVqfKmY6rsTLXzNgk
A Patreon token pasted into Add feed, or a creator link without
&show=, becomes a folder of that creator's shows, found through
Patreon's web API and kept in step like a subscribed OPML (sync_group,
split out of sync_opml). A creator already read as one feed is split
too: each show takes over the files and read state it held
(Db::adopt). A creator with one show stays a plain feed.
Filter verdicts are judged again every scan, so turning on Allow
explicit brings skipped items back. Add feed has an explicit box.
Feeds in a group follow your settings on the group, as its dialog
said. A new feed no longer takes the id of a removed one at a
different URL and shows its old items. See CHANGELOG.md [Unreleased]
and docs/history.md.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wi22VSVrkAvqNj61eqHsm9
- The import screen has a file picker beside the paste box. The page
reads the file, checks it looks like OPML before sending, and clears
the picker when it is refused and after it is imported. The file is
sent as text and never written to disk on the server.
- The server parses the OPML before touching anything and answers 400
"that is not an OPML file" (was a 500). subscribe_opml takes a parsed
document, so ipx import also refuses a non-OPML file by name.
- Tests: Settings' Export OPML download and paste import; uploading an
RSS file (refused) and a real OPML; the server's 400; the admin's
export round-tripped into a second account; ipx import/export in a
scratch config.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016HdTEWQNrzyFULijigkmMn
Import predated accounts: it only added URLs missing from config.toml
and subscribed nobody. Importing another account's export did nothing
("Imported 0 feed(s)"), and a genuinely new feed had no subscriber, so
it was never scanned. Web and CLI import now share subscribe_opml,
which subscribes the caller (the CLI: the first admin) to every feed in
the file and reports new vs already-subscribed.
Export wrote the whole catalogue to anyone signed in, including other
people's private feed URLs. It now lists only your own subscriptions.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0173mGu6rK18Ne7UGTwAaVJV
Read, starred and position move to entry_state; subscriptions carry each
person's keywords, auto-download, explicit and per-scan limit. The feed
list and unread counts are per person, and the existing library is
adopted by the admin on first start.
The feed URL, folder and schedule stay shared and admin-only: one file
serves everyone, so they describe the file rather than a preference.
Scanning merges subscribers' wants -- anyone wanting an item is enough --
via merge_policy, which is pure and tested.
Also: the test fixture wiped its data directory from every Playwright
worker, deleting the database out from under the running daemon.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
epAction's redraw closure called itself when handed a row, so Mark read
recursed until the stack blew; it now swaps that row in place. Opening an
item also marks it read, redrawn where it stands so nothing vanishes from
under the pointer on the Unread tab.
Step A of multi-user: users and sessions tables, Argon2id, a session
cookie, ipx user subcommands, and a trusted proxy header for Cloudflare
Zero Trust -- honoured only from a trusted_proxies address. The shared
token still works and is the admin. A new database starts with
admin/ipodderx.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
Deleting entries during a cleanup left each feed's ETag in place, so the
rescan got 304s, skipped parsing, and 57 feeds stayed empty until a
publisher happened to change something. A 304 while the feed holds zero
entries means the validator has outlived the data, so the daemon drops it
and asks again.
"not a wanted media type" was jargon, and storing it in last_error painted
an ordinary filter decision red. It reads "not audio or video" now, and a
reason is only shown as an error when the state actually is one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
Blog feeds put each article's header image in an <enclosure>, so a text
feed read as a podcast full of episodes: 149 images, 74 MB across 11
feeds. media_types defaults to audio and video, with a per-feed override.
Such enclosures stay listed and stay downloadable by hand; the row names
what it is rather than saying "skipped". An unknown type is allowed, since
the real type is only known after downloading, and a torrent is allowed as
a container judged once unpacked.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
The log view splits into All / Daemon I/O / Scans / HTTP. Daemon I/O is
the control protocol itself, logged where every command funnels through
so it covers socket clients, the CLI and the web UI alike. stderr and the
in-app buffer now have separate filters, so the UI can keep debug detail
the terminal should not carry.
Playwright drives a real browser against a daemon on fixture feeds. Eight
tests, each mapping to a bug that reached a user -- the Rust tests and the
stub-DOM smoke test cannot see a wrong selector or a dead handler.
It immediately found one: OPML folders rendered expanded by default,
because the code stored closed groups, so any folder never toggled counted
as open.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
Moving OPML feeds into the database left several call sites still
searching config.toml only, so anything inside a subscription looked
unsubscribed: Download failed outright, status and the startup line
counted 3 feeds instead of 85, add could duplicate or collide with a
derived feed, and rm could not remove one.
The first grep for this missed the failing call because the method chain
spans lines; searching with newlines collapsed found all of them.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
Writing 82 derived feeds into a hand-edited config.toml made it
unreadable. The OPML is the source of truth, so its feeds are re-derived
each scan and held in the database, inheriting the subscription's
settings; editing one promotes it to a real entry. A migration moves
existing children out -- 611 lines to 38 -- keeping all entries and files.
max_new_per_check defaulted to unlimited, so subscribing to an OPML of 82
feeds pulled whole back catalogues. It now defaults to 3 via [general],
capping every feed that does not set its own, and the pending queue orders
by publish date so a cap of 3 means the three newest.
Scans and downloads travelled as socket events only, so the log view
showed no daemon activity. They are mirrored into tracing, with routine
skips at debug -- at 82 feeds those alone would flush the buffer.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
A feed whose body sniffs as OPML is treated as a subscription list and
re-read on every scan, as iPodderX did. Listed feeds become real config
entries grouped under it, inherit its settings, land in one nested folder,
and are scanned in the same run.
When a feed leaves the OPML: removed if nothing was downloaded, kept and
flagged otherwise, so a downloaded file is never orphaned.
folder_for sanitized the whole folder string and would have flattened the
nesting; each segment is sanitized separately now, and a traversal still
cannot escape the download directory. Db::memory() also runs migrate(),
which it did not, so a migration-only column passed tests while missing in
production.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
The Log button shows the running daemon live: feed scans, downloads,
torrents and every HTTP request. It reads a ring buffer filled by a
tracing layer rather than tailing a file, so it works under Docker where
logs go to stdout. The access-log middleware skips /api/logs, or the
panel's poll would log itself forever.
Detached torrents could leave a row stuck in 'downloading' across a
restart, where nothing would ever revisit it; those are requeued at
startup.
Dockerfile, entrypoint and compose: 114 MB runtime, config bound to
0.0.0.0 on first run since container loopback is unreachable, drops to
PUID:PGID for Unraid, and a healthcheck that goes through the control
socket so a wedged worker reads as unhealthy.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
A torrent ran inline on the single sequential worker, so it blocked every
feed scan, HTTP download and status command behind it -- for up to
stall_mins waiting on metadata, and for up to seed_time_mins seeding after
finishing. A live daemon was wedged with 47 pending torrents and would not
answer a status command for 15s.
spawn_torrent detaches the job behind a 2-permit semaphore and marks the
row 'downloading' so a rescan cannot queue it twice. One-shot CLI runs
stay inline, or the process would exit mid-download.
Torrents themselves verified working: a 755 MB Debian netinst downloaded
to completion against a real swarm.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
"Check feeds every" becomes a number plus a unit dropdown in both global
and per-feed settings; parse_interval gained weeks to back it. The
per-feed dropdown can select the global default, clearing the override.
Fixes progress painting every pending row: the event carried no enclosure
id, so the handler had nothing to target and set the width on all of them.
Adding a feed looked like it was downloading everything. Progress,
DownloadDone and DownloadError now carry the enclosure id.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AdXho5tTkjFLeUXKbEjKBh
general.schedule and feeds.<id>.schedule take "every 30m", "4h", "1d" or
bare minutes. The legacy interval_mins is still read. An explicit per-feed
schedule wins over the publisher's ttl; without one, ttl still raises the
interval when they ask to be polled less often.
Fixes two bugs found while testing it:
null never cleared a field. serde maps JSON null onto the outer None of an
Option<Option<T>>, so "clear this" was indistinguishable from "not
supplied" and every clear silently no-opped with a 204.
The daemon ignored SIGTERM while working. select! races branches only at
selection time, so a signal queued behind an in-flight download and the
process had to be SIGKILLed. The stop signal now cancels work in progress:
SIGTERM mid-download exits in 1s.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
Rewrites the page around a persistent player (speed, seek, resume,
MediaSession, keyboard shortcuts), artwork, filter tabs, episode search,
pagination and live progress, with modals and toasts replacing prompt()
and a status line.
Backend gains the metadata that makes that possible: feed and episode
artwork, durations, season/episode numbers and playback position, plus
filters, search, totals, mark-all-read, download-latest and OPML over
HTTP. Schema changes arrive through a real migration, since CREATE TABLE
IF NOT EXISTS does nothing to an installed database.
Fixes filtering, which returned 500 whenever no search term was given:
the search clause was dropped while its parameter was still bound.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
POST /api/enclosures/{id}/download requeued the row and asked for a
normal scan, but a scan takes the lowest-id pending rows up to
max_new_per_check. With a large backlog and a small cap the requested
row was never a candidate, so other episodes downloaded while it stayed
pending.
A queue expresses what is outstanding, not what was asked for. Download
is now its own command that fetches one specific enclosure immediately,
ignoring queue order and the per-scan cap, still via the single worker.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
axum served from inside the daemon so it reads SQLite and the event bus
directly: browse feeds, read show notes, play with seeking, download and
delete files, mark read/flag, and edit feed settings.
Config is now hot-reloadable (Ctx.cfg behind RwLock<Arc<Config>>), so UI
edits apply without a daemon restart. Access is a shared token minted from
/dev/urandom, carried in a cookie because an <audio> element cannot send
headers. Show notes are untrusted feed HTML and are sanitized with ammonia
server-side.
read/flagged finally have a writer, which retention has needed since it
started ordering by them.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
librqbit replaces the vendored BitTorrent 4.2.1 tree. Torrents download
in place because seeding serves the files it downloaded, so the planned
stage-then-move would have broken it. The stall budget now also covers
magnet metadata resolution, which otherwise never returns against a dead
swarm and wedged the scan.
Adds add/rm/import/export, tracing setup, systemd units and README.
A successful swarm download is unverified: no reachable peers here.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
Unix-socket JSON-lines protocol replacing the printMSG sentinels, with
an Emitter so scan code is agnostic about whether a terminal or a UI is
watching. Daemon serves clients and runs a TTL-aware scheduler; commands
funnel through one worker so scans cannot overlap, and the CLI proxies to
a running daemon rather than competing with it.
The pre-fetch retention sweep was emitting the terminal ReapDone event,
which would have ended a client's read before the scan began.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
Oldest-first reaper with the original's 50 MB headroom pad, plus a
reconcile pass for files deleted by hand and pruning of stale entries.
The Python meant to reap only read, unflagged episodes but a missing
import and a typo made that filter throw on every candidate. Requiring
read=1 would be equally dead headless, so flagged is the keep-forever
marker and read only decides ordering.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
Streaming download with progress, content sniffing that replaces the
long-dead detectFileType(), HTML-body rejection, and placement by rename
from an incomplete dir on the same filesystem.
Filters are applied once at discovery and recorded in enclosures.state,
so the download queue is the table rather than the parse result and
max_new_per_check defers work instead of dropping it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
Conditional GET plus an RSS-first, Atom-fallback parser normalising both
into one entry model, with iTunes explicit and ttl handling carried over
from the Python. Enclosures are recorded as pending; nothing downloads yet.
GUID falls back guid -> link -> enclosure url -> title rather than hashing
the description as the original did.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
config.rs replaces iPXSettings.py and feeds.plist; db.rs replaces the
per-feed .ipxd plists, history.dat and qmcache.dat. enclosures.url is
UNIQUE, which is the dedupe key the old pickle history provided.
ipx list is the first working subcommand.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
Rust rewrite of ipodderx-core: cargo project, pinned dependency set,
MIT license carrying the original 2010 copyright, README lineage note,
and PROGRESS.md as the resumable build record.
Dependency tree is rustls-only: librqbit's default features pull in
native-tls and an OpenSSL sha1 backend, so it is built with rust-tls.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RPyeapneuXrCdojsaiXGbe
