Nexus 0.25.0
Released 1 August 2026. The current release is 1.24.0. Download it · Every version
Fixed: torn-off windows open at a readable size
Pop out the CW band map and the type was tiny — and stayed tiny at every size the window could be dragged to, docked or free. Every torn-off window sized itself by asking whether the full Operate cockpit fit inside it. A 420-pixel-wide band-map strip never fits a cockpit, so the answer was always "no" and every pop-out opened at the smallest scale the app has, permanently.
Each pop-out is now measured against its own contents instead. The band map, the waterfall strip, the Needed board, Connect, DXpeditions, Memories, Satellites and the Field Day scoreboard all open at full size, and shrink only when you actually squash the window down toward its minimum — where shrinking is the right answer. The Operate pop-out is unchanged: it really does host the cockpit, so the cockpit is the right question for it. The main window is unchanged at every size.
If you have pinned a scale in Settings, that pin now applies in a pop-out as far as the window can take it, instead of being crushed to the smallest step. A pin is still never rewritten by a pop-out, and still never enlarged past what the window can show.
The Needed board no longer carries its 25% font compensation — it existed only to offset the scale bug, and the board now reads larger than before without it.
Added: the satellite catalog ships with the app
A new install used to start with nothing and go looking for orbital elements. If the element mirror could not be reached, the fallback fetch brought back CelesTrak's amateur group — 97 objects, no status information — so the Satellites screen listed a fraction of the birds and could not say a word about any of them: nothing marked dead, re-entered or not yet launched, nothing marked alive but silent, no row explaining a bird it holds no elements for, and a starred bird that stopped working simply vanished from the list with the star still set.
The installer now carries a snapshot of the full catalog — 430 birds, 367 of them with elements — so the Satellites screen is complete from the first launch, with no network at all. The snapshot is a floor, never a ceiling: it fills in what your copy is missing and never replaces newer elements with its own, never overwrites the statuses you already have, and never counts as a fetch, so Nexus still refreshes from the mirror at the first opportunity exactly as before. Upgrading from an earlier version keeps everything already cached and gains the catalog beside it. Elements from the snapshot age like any other: past 30 days they stop being used, because SGP4 accuracy is gone by then and a position would be a guess.
Satellite population, names and status are derived from the SatNOGS DB (CC BY-SA 4.0); orbital elements are courtesy of CelesTrak and the SatNOGS TLE API.
Fixed: SSTV receives pictures without being told to, and says what it is hearing
You could hear a picture coming in on the speaker, watch it march up the waterfall, and still get nothing — no image, no progress, no explanation. Opening the SSTV screen now starts the receiver. Until now it had to be armed by hand every session, and until it was, the decoder was handed no audio at all, so the ordinary way to use SSTV (open the screen, tune 14.230, wait) decoded nothing. The waterfall and the decoder are fed by two different paths, which is why a live waterfall never proved the decoder was being fed — and why the failure was invisible. Stopping the receiver yourself is still respected: it stays stopped for the rest of the session, and nothing here can transmit.
The line under the waterfall now states what the receiver actually hears rather than one fixed hint. It tells apart a stopped receiver, a capture device delivering nothing, an input that is alive but silent, a picture in progress, and — new — a station sending in a mode Nexus cannot decode, which previously looked exactly like a dead band because it was only ever written to a console log. Hearing the signal on the speaker says nothing about what the app is capturing, and the screen now says so instead of leaving you to guess.
More SSTV frequencies, and the right one. The band list gained 160, 17, 12 and 6 m, the European 40 m calling frequency, and the 20 m overflow channels 14.233 and 14.236 — the last of which is where the original report came from, and which existed nowhere in the app. The on-screen hint now names the calling frequency for the band you are on instead of reciting 14.230 and 145.800 whatever you are tuned to. While a picture is coming in, the caption says how long that mode takes on the air, so a 110-second Scottie 1 no longer looks like a hang.
Three more things that came out of reviewing the fix. An ISS pass ending no longer leaves SSTV switched off for the rest of the session — the automatic disarm at loss of signal was indistinguishable from you pressing Stop, so anyone using the ISS auto-arm option would have lost SSTV decoding from their first pass of the day onward. The new status line no longer blames your sound card in the second before the decoder has reported anything; until it has heard something it says so plainly. And 28.680 is no longer described as Technician-accessible: US Technicians have 10 m images only on 28.300–28.500.
Fixed: the map stopped throwing away its canvas every second
On a wide display Nexus's memory use oscillated by ~20 MB once a second — visible in Task Manager as a steady sawtooth. The map's draw pass reassigned the canvas size on every run, and assigning that size discards the whole image buffer and allocates a fresh one even when nothing about the size changed. Since the draw pass runs on a one-second pulse whenever an animated layer is showing (band openings and the heat layer are on by default), a full-window buffer — about 20 MB at 3440x1440 — was thrown away and rebuilt every tick. Nothing leaked; the memory was reclaimed each time. But it was work for no result, and on a slower machine that kind of churn is felt as stutter rather than seen in a graph.
The map now resizes its canvas only when the size genuinely changes, matching what the waterfall has always done. The flare overlay got the same treatment (it was rebuilding a full-window buffer on every zoom and drag), as did the small spectrum display in Connect and Settings, which was doing it eight times a second. A test now fails the build if any drawing surface goes back to resizing on every frame.
Added: satellite operating without the box — and passes that know their worth
No rotator? Tracking still works. Arming a pass no longer needs a rotator: the pass clock, the sky dome and Doppler all run, and only the pointing is skipped — the handheld-antenna operator gets everything but the mast, including where to swing the antenna. The tracking badge, the readiness rail and the rotor strip say exactly which surfaces a track actually drives — rotor, dial, both, or neither — and a track driving neither says "pass timing only" instead of implying more. While a rotor-less pass holds the dial, an ownership chip appears in the cockpit header naming the bird, with a stop button that hands the dial straight back; stopping a live track always releases the dial.
The uplink's sideband is set with its frequency. On an inverting bird you listen USB and transmit LSB. Nexus now puts the TX (split) VFO in the right sideband along with its frequency while a pass owns the uplink — commanded once per answer, only when the two legs genuinely differ, and only onto the satellite's own uplink: a terrestrial pile-up split worked while a transponder is held keeps its own sideband. Reach for the rig's mode knob yourself and Nexus stands down for the rest of the pass rather than fighting you. If the rig refuses the command, the status line says which sideband to set by hand — a wrong uplink sideband sounds exactly like nobody answering. The Doppler readout and the transponder chooser show what is actually being commanded, and say so plainly when nothing is.
Passes ranked by what they would earn you. The schedule and the "Next up" strip now say what each pass is worth: the grids you have never worked via satellite and the entities you have never worked at all, wherever the pass's footprint crosses them — in the same need-chip language as the Needed board, with the sample squares on the pass timeline. "Needed" is a column you click to sort by; the default order stays soonest-first, and a bird SatNOGS reports dead still shows its dead tag here. Satellite VUCC is now its own card on the Awards screen — ARRL counts a satellite contact toward that category only — so the terrestrial VUCC card counts terrestrial grids, and a satellite QSO no longer silences a NEW GRID call-out for a band slot that is genuinely still open.
One click to work a pass. ▶ Work this pass opens the bird, picks a workable transponder for you — never a beacon, never one reported dead, and never overriding a "None — leave the dial to me" you said for that bird — and arms the track. A readiness rail under the bird's name shows the whole chain (pass, rotor, transponder, Doppler), each gate fixable where you are standing. The transponder chooser is a card list beside the tuning instruments with dead entries folded behind one line, and the sky dome, timeline, chooser and passband strip sit together above the globe. What the rail and chooser show is what the engine actually holds: when a pass ends or is stopped, the hold is released and the display follows.
Picking a transponder tunes the radio. Choosing a bird — by hand or through ▶ Work this pass — puts the rig on that transponder's downlink there and then, with the uplink on the transmit VFO your mapping calls for. No waiting for AOS: the pass takes over from where the pick left the radio. The click is consent for the dial only, exactly as clicking a spot or a repeater favourite already is; every transmit gate is unchanged, and "None — leave the dial to me" still means the radio does not move. If Doppler is switched off, your mapping is None, or the rig cannot reach the band, the transponder is still held and the section tells you plainly that nothing was tuned.
The Satellites section says which radio it drives. A dual-radio station no longer has to guess. Picking a transponder routes on band and mode class the same way a repeater tune does, so a VHF/UHF bird reaches the VHF/UHF rig even when you were sitting on HF, and an FM bird follows your FM routing rule while a linear bird follows your SSB one. A line under the held bird's name names the rig, the band and the class it routed on, and the frequencies it actually wrote — with a pin to hold the current radio when you want to override the routing.
A routing rule can now say "Satellite". Asked for from the field: mode-class routing sends a packet bird through your FM & APRS rule — right by the rules, wrong for the shack where APRS lives on one rig and the satellite antennas on another. The rule editor's mode dropdown gains Satellite: a rule so designated is checked before the mode rules, but only when a satellite tune is asking — picking a transponder finds it, a terrestrial FM, APRS or FT8 tune never does. Without a Satellite rule nothing changes: satellite picks keep routing on band and mode class exactly as before. One caution: builds before this one don't know the designation and read a Satellite rule as a plain terrestrial rule — with no band or mode set, one that catches every tune — so if you ever roll back, delete the Satellite rule first.
No more FT8 frequencies under Satellites. The band dropdown at the top of the window is fed the digital watering holes, which meant it offered 14.074 beside a bird on 435 MHz. Satellites now owns its own frequency surfaces, like Phone, CW, RTTY, SSTV and APRS already do.
Orbital elements stay current — and say so when they can't. Elements now refresh in the background from the project's mirror of CelesTrak's amateur list, held in one shared on-disk snapshot: no more re-fetch on every launch, nothing ever waits on the network, and CelesTrak itself is asked directly only as a narrow, rate-respecting fallback. The stale badge is now a refresh button, and Settings gains an Orbital elements block with Update now and Import from file — the path for offline shacks and brand-new launches. Update now means now: if the mirror can't deliver when you press it, the same attempt goes straight to CelesTrak instead of telling you to come back tomorrow — still honoring CelesTrak's update cycle, and never after CelesTrak has said stop. And the answer speaks operator, not protocol: a fetch that fell through to CelesTrak says so with the bird count, an unreachable mirror over current elements says your elements are fine (the mirror goes live with the next site release), and a real failure says what failed and what to do next — no more "HTTP 404" as the answer to a button press, with the raw error riding the tooltip for troubleshooting. Element age is enforced where it matters: past 14 days arming a pass asks first; past 30 days the detail pane, tracking and the SSTV auto-arm refuse plainly, naming the bird and the age, instead of pointing the antenna with a fiction. A pass keeps the elements it armed with (their age shown on the readiness rail), a bad or empty download never replaces a good cache, and a CelesTrak rename no longer orphans your starred birds, alarms or schedule — Nexus remembers the catalog number behind each name.
Connect's satellite layer follows your stars. Turning on satellites in Connect now shows the birds you have starred in the Satellites section — on the 2-D map, the 3-D globe and the Passes pane alike — instead of the whole amateur catalog. A small ★/All chip flips between your birds and everything — on the Passes pane and beside Satellites in the map and globe Layers panels — one choice all three surfaces share and apply the moment you flip it, and with no stars set the full sky still shows so a fresh install is never empty. If every starred bird has dropped out of the current elements, the map and globe say so rather than drawing a silently blank sky. Every bird on the map and globe now carries its designation ("ISS", "RS-44") beside its satellite icon, and map labels that would overprint each other in a cluster shuffle apart so both names stay readable. Stars recognized by catalog number too, so an upstream rename never drops a bird from your filtered sky.
The bird list is the birds that are actually up there — and it says when one stops being one. Nexus used to show one Celestrak list of 97 objects, and measured against AMSAT's live reports roughly 60 of those had not been heard in a month, while eight birds heard on the air that same day were missing from it entirely (IO-86, Foresail-1p, SAKHACUBE, QMR-KWT 2, Ten-Koh 2, Marina and two more). The list is now built the other way round: start from the satellites the SatNOGS database says carry an amateur transmitter, then go and find current orbital elements for the ones that are actually workable — from the two Celestrak groups and the SatNOGS element service together. That is 430 birds listed and 367 with live elements, against 97 before: 274 of them are birds the old Celestrak group never carried at all. Birds catalogued under a placeholder number until Space Command assigns a real one (normal for the first months after launch) now appear under one entry instead of two or none.
Every bird carries its status with it, so the Birds list, the map and the globe can show what only the Satellites schedule could show before — and a bird that stops being workable keeps its row and says why: dead, re-entered, pre-launch, or alive but silent when nothing amateur is transmitting on it any more. A starred bird that dies no longer disappears out from under your ★; it stays in the list, wearing the reason, until you take the star off yourself. Birds nobody holds current elements for are reported the same way — no elements at all, elements too old to trust, or an orbit decayed past the point where a position can be computed. The Connect Passes pane names those birds under the passes, and searching the Birds list reaches the whole catalog, elements or not, so a bird you unstar is always a bird you can find again. A bird that re-entered more than six months ago drops off: by then it is history, not a catalog.
And your first ★ birds are set for you — once. A fresh install used to open Satellites on an empty schedule beside a list of hundreds of birds, with nothing to say which are worth chasing. Nexus now stars the ten most workable active birds over your grid the first time it can: alive, carrying a live amateur transmitter, with current elements, and flying real passes over you — ranked by how many of those passes you get and how high they climb. A line at the top of the section says it happened and disappears when you dismiss it, and every star it set is one you turn off where you are standing. It happens once. Clear the whole set and you get an empty sky, not the same ten back next launch; a station that already has favourites is never seeded over, and neither is one that deliberately cleared theirs — including one that cleared them on an older version. With no grid square set, or before the bird catalog has landed, nothing is starred at all: a guess is worse than waiting.
The mirror that feeds all this refreshes every six hours and refuses to publish at all if any of its sanity checks fail, so a bad upstream day leaves your last good list in place rather than replacing it with a short one. If the mirror itself is unreachable for a day, Nexus falls back to fetching Celestrak's amateur group directly — a shorter list, and every bird it does not carry keeps its row, marked as having no current elements.
Fixed: the pick reaches the radio — and the IC-9700's uplink goes where it can transmit
A dual-radio pick finds the satellite rig even with nothing configured. Field report: on an FTdx10 + IC-9700 station, clicking a transponder moved nothing — no routing rule named the 9700 for the band, neither rig listed bands, and the pick stopped at "this radio doesn't cover 435 MHz" while the 9700 sat idle. When no rule or band list answers but the active rig is known unable to reach the downlink and exactly one other radio exists, the pick now hands the QSY to that radio. With several candidates and nothing naming one, the section refuses and says exactly what to configure instead of guessing which rig to move. A radio whose band list explicitly leaves the downlink's band out is never the fallback — that list is your word that the rig doesn't take the band, and the pick refuses rather than switching to it just to be turned down.
A packet bird is FM to the radio — and routes like it. Field report: on the same station, picking the ISS APRS digipeater (145.825 up and down) selected the HF rig and called the bird SSB. The mode class behind routing recognised only the literal names "FM"/"FMN", so every packet mode SatNOGS uses — AFSK, FSK, GMSK and the rest of the family — fell through to the SSB class: the FM routing rule never matched, and USB would have been commanded on an FM channel. One mode-name map now classifies the whole packet family as FM everywhere the class is consulted — routing, the commanded rig mode, and the uplink's sideband — so the digipeater reaches the VHF/UHF rig, in FM. Linear birds (the RS-44 class) behave exactly as before, and a mode name the map has never seen still reads SSB, as it always did.
A simplex bird rides one dial. 145.825 up and down is one channel, not a cross-band pair: the pick no longer writes a split for it, nothing engages the rig's satellite mode, and mid-pass Doppler holds the dial on the published frequency — the two legs' corrections are equal and opposite, so steering the one dial to either leg lands the other twice as far off, outside an FM passband. Parking on the channel is what every 145.825 operator does by hand, and now the tracker does the same. The radio line shows the one frequency once, and the Doppler row says the two legs share the dial instead of printing the VFO mapping twice. Cross-band FM channels (SO-50's 145.850 ↑ / 436.795 ↓) keep both dials steered as before.
Cross-band uplink on the IC-9700 rides satellite mode. The uplink used to be written as an A/B split — which on a dual-band rig lands in the downlink's own band and goes nowhere. Under a Main = downlink / Sub = uplink mapping, Nexus now engages the rig's satellite mode, writes the uplink (and its sideband, on an inverting bird) into the Sub band with a read-back check, and hands the tuning selection straight back to Main, so the dial, the scope and every poll keep the downlink. Releasing the split releases satellite mode; switch it off on the front panel yourself and Nexus reports that instead of re-engaging over you. A rig without a Sub band refuses the mapping honestly, Main = uplink / Sub = downlink is refused as undrivable (satellite mode always transmits on Sub), and A/B rigs keep the existing split behaviour byte for byte. An ordinary A/B split commanded while satellite mode is engaged — WSJT-X setting up split for a digital over mid-pass — releases satellite mode first, so its transmit dial can never land in the Sub band and go out on the downlink. And the sequence trusts nothing it cannot confirm: an engage whose confirming read-back is lost is backed out rather than left half-set, a hand-back to Main the rig refuses is remembered and re-asserted before the next tuning write or dial poll, and a rig that will not leave satellite mode is reported in the status line instead of being shown as simplex.
The radio line reports what the rig acknowledged, not what was computed. The binding under the held bird used to print both frequencies the moment you clicked — including when nothing had been sent. Each leg now shows as still tuning ("435.640 ↓ …") until the radio actually accepts it, the dot fills only when every requested leg is confirmed on the wire, and a leg the rig refuses turns into the reason in plain words. A refused pick no longer flashes a green "Working …" toast — the toast tells the truth the read-back found.
Clicking a transponder no longer freezes the app (0.24.3 tester build). Field report: on the IC-9700 under the native CI-V daemon, picking a bird and clicking a transponder frequency froze the whole window until Windows killed it, every time. The radio loop deadlocked on itself while applying the uplink split — it took the engine's own lock a second time while still holding it — and every part of the UI then queued behind that lock forever. The same wedge caught terrestrial pile-up splits ("UP 5" spots) and every mid-pass Doppler correction. The apply now releases the lock before talking to the rig, and a liveness test drives the real pick against a simulated 9700 under a watchdog so this class of freeze fails the build instead of the operator. And because the rig conversation now runs unlocked, a split the operator requests during it is safe: a rejection names the dial it was rejecting and resolves only that request, so a fresh "UP 5" click in that window still applies on the next cycle instead of being silently dropped.
Fixed: RTTY no longer freezes the window when the copy gets rough
Field report: "getting some application hangs on RTTY." The decoded-text pane fades each character by how confidently the demodulator copied it. That confidence is a continuous measurement, so on clean copy the whole transcript was one solid block — but as conditions degraded it crossed the fade thresholds constantly, and the pane ended up drawing the transcript one character at a time: up to four thousand separate pieces, redrawn twice a second, forever, because the transcript keeps scrolling. Bench-testing on a strong signal could never show it; a noisy band or an armed decoder listening to empty air brought it on, which is why it came and went. The pane now draws at most a couple of hundred pieces no matter how bad the copy is — the same "cap the feed" rule the decode history already follows. Every decoded character still prints, and the fade is still scored character by character for any transcript under that ceiling, which is every transcript you can actually read. Only when the copy is breaking up badly enough to blow the ceiling does the fade get scored over short stretches instead of single characters; from that point a brief marginal burst can be averaged in with the good copy around it and shown solid. That is a deliberate trade, and it is only ever charged on copy already too broken to trust.
A stalled radio can no longer stop the window from responding. Nexus talks to the rig over a serial link that can take up to 2.5 seconds to answer, and while it waits it holds the lock that the rest of the app needs to read anything. Those reads were running on the same thread Windows uses to paint and respond to clicks, so a slow or wedged rig could park the window long enough for Windows to declare it "not responding" — with nothing in any log, since nothing had actually crashed. Those reads now run off the painting thread: a slow rig costs a late reading, not a dead window. The long jobs got stronger treatment — a LoTW, eQSL or QRZ sync can sit on the network for a minute, and those now run on a pool set aside for waiting, so a slow QSL server can't hold up the rest of the app either. A build-time check keeps it that way; it is keyed on which commands can reach the shared radio state, so a new one can't slip through by reaching for it in a different way.
One misbehaving companion app can no longer freeze the radio loop. A logger that jams the WSJT-X control port with stop-transmit commands was previously handled all in one pass, each one costing a full round trip to the rig — with the app-wide lock held the whole time. Nexus now handles a bounded batch per cycle; the rest waits for the next cycle a few milliseconds later, so nothing is dropped.
And the class of bug behind the recent satellite-picker hang is now caught at build time. That freeze came from a Rust pattern where a lock is accidentally kept open longer than it looks — invisible in review, with no symptom until the app is already frozen. The build now rejects the pattern in the command layer where it does the damage, and flags it everywhere else. Two spots in the FlexRadio streaming code that matched it were rewritten (both harmless today, neither harmless to leave).
Fixed: the Openings Log square works — and a broken screen can no longer black out the app
Turning on the Openings Log square blacked out the whole app (0.24.6 tester build). Field report: assigning "Openings Log" to a Connect slot turned the entire window black — rail, top bar, everything — and the black screen came back on every launch, because both the slot assignment and the openings journal itself outlive the session (the journal even outlives a wiped browser profile). The pane paints in two steps — an empty first paint while the log loads, then the real render — and its sort state was declared on the wrong side of the "nothing yet" exit, so the two paints disagreed about the component's shape and React tore down the entire window. The sort state now sits above that exit, the pane renders whatever the journal holds — including rows written by older builds with fields missing — and a build-time sweep now fails on this pattern anywhere in the UI, so the class is closed, not just this instance.
A broken screen can no longer black out the app. A section that crashes while drawing now shows an error panel in its place — what crashed, the technical details ready to copy into a bug report, and a button back to a known-good section — while the navigation rail and top bar stay alive around it. Pop-out windows carry the same net. And the section the app reopens on is checked against what this build can actually render before it is restored, so a saved id from an older or newer version can never wedge startup into a dead screen.
Added
-
The recall card's picture opens their QRZ page.
Click the callbook photo on the station you are working — or the initials circle when they have no photo — and QRZ opens in your browser, so you can read the page while you are still in the QSO. The circle is a link only once the callbook has actually resolved the call; before that it is the same plain badge it always was. Clicking it does not move the cursor out of the log field you are part-way through typing, and it is reachable from the keyboard like any other control.
-
CQ+73 filter chip
in Band Activity (a tester request): the CQ view plus RR73/73 signoffs — a signoff means that frequency is about to free up. Sits between CQ and To me; the plain CQ chip is unchanged. Detection rides the message parser, so a
DM73grid or anRRRroger never counts as a signoff.
Fixed
-
TX Off follows WSJT-X's Enable-Tx contract in the FT modes.
Turning TX Off lets a transmission already in flight finish its over, and the next cycle simply doesn't start; turning it back on before the next cycle transmits normally. Stop TX is unchanged — it halts immediately, mid-over. Previously both controls cut the transmission on the spot. Only the FT-style cycle gets to finish: in CW, RTTY, SSTV, the voice keyer and APRS, TX Off still stops the over on the spot, and the cockpit Stop buttons now unkey even in the last fraction of a second of an over instead of leaving the rig keyed to the tail.
-
The Band Activity filter chips wrap onto a second line instead of running off the edge of a narrow rail.
The chip row could not wrap, which made it a single unbreakable block: with the rail dragged to its narrowest the last chips sat past the panel edge, and nothing in that column scrolls sideways, so there was no way to reach them. Adding a seventh chip turned that from marginal into certain.
Every version
- 1.24.0 11 Nov 2026
- 1.23.0 5 Nov 2026
- 1.22.0 1 Nov 2026
- 1.21.0 26 Oct 2026
- 1.20.0 22 Oct 2026
- 1.19.0 16 Oct 2026
- 1.18.0 12 Oct 2026
- 1.17.0 7 Oct 2026
- 1.16.0 3 Oct 2026
- 1.15.0 26 Sept 2026
- 1.14.0 20 Sept 2026
- 1.13.0 16 Sept 2026
- 1.12.0 14 Sept 2026
- 1.11.1 8 Sept 2026
- 1.10.3 4 Sept 2026
- 1.10.2 3 Sept 2026
- 1.10.1 1 Sept 2026
- 1.10.0 31 Aug 2026
- 1.9.2 29 Aug 2026
- 1.9.1 26 Aug 2026
- 1.9.0 25 Aug 2026
- 1.8.1 23 Aug 2026
- 1.8.0 23 Aug 2026
- 1.7.6 21 Aug 2026
- 1.7.5 20 Aug 2026
- 1.7.0 18 Aug 2026
- 1.6.1 17 Aug 2026
- 1.6.0 16 Aug 2026
- 1.5.0 16 Aug 2026
- 1.4.0 15 Aug 2026
- 1.3.0 13 Aug 2026
- 1.2.0 10 Aug 2026
- 1.1.0 10 Aug 2026
- 1.0.5 8 Aug 2026
- 1.0.2 6 Aug 2026
- 1.0.1 5 Aug 2026
- 1.0.0 4 Aug 2026
- 0.28.2 3 Aug 2026
- 0.28.0 3 Aug 2026
- 0.27.0 2 Aug 2026
- 0.25.0 1 Aug 2026
- 0.24.0 31 Jul 2026
- 0.21.5 29 Jul 2026
- 0.21.0 29 Jul 2026
- 0.20.0 28 Jul 2026
- 0.19.7 27 Jul 2026
- 0.19.6 26 Jul 2026
- 0.19.4 26 Jul 2026
- 0.18.0 25 Jul 2026
- 0.17.22 25 Jul 2026
- 0.17.21 25 Jul 2026
- 0.17.20 25 Jul 2026
- 0.17.19 25 Jul 2026
- 0.17.18 25 Jul 2026
- 0.17.17 25 Jul 2026
- 0.17.16 25 Jul 2026
- 0.17.15 25 Jul 2026
- 0.17.14 25 Jul 2026
- 0.17.13 25 Jul 2026
- 0.17.12 25 Jul 2026
- 0.17.11 25 Jul 2026
- 0.17.6 25 Jul 2026
- 0.17.5 25 Jul 2026
- 0.17.4 25 Jul 2026
- 0.17.3 25 Jul 2026
- 0.17.2 24 Jul 2026
- 0.17.1 24 Jul 2026
- 0.16.4 24 Jul 2026
- 0.16.3 24 Jul 2026
- 0.16.2 24 Jul 2026
- 0.16.1 24 Jul 2026
- 0.16.0 24 Jul 2026
- 0.15.24 24 Jul 2026
- 0.15.23 24 Jul 2026
- 0.15.22 24 Jul 2026
- 0.15.21 24 Jul 2026
- 0.15.20 24 Jul 2026
- 0.15.18 24 Jul 2026
- 0.15.17 24 Jul 2026
- 0.15.16 23 Jul 2026
- 0.15.15 23 Jul 2026
- 0.15.14 23 Jul 2026
- 0.15.1 22 Jul 2026
- 0.15.0 21 Jul 2026
- 0.14.0 21 Jul 2026
- 0.13.0 19 Jul 2026
- 0.12.0 18 Jul 2026
- 0.11.1 18 Jul 2026
- 0.11.0 18 Jul 2026
- 0.10.0 17 Jul 2026
- 0.9.7 17 Jul 2026
- 0.9.6 16 Jul 2026
- 0.9.5 16 Jul 2026
- 0.9.4 16 Jul 2026
- 0.9.3 16 Jul 2026
- 0.9.2 15 Jul 2026
- 0.9.1 15 Jul 2026
- 0.9.0 15 Jul 2026
- 0.8.9 15 Jul 2026
- 0.8.8 14 Jul 2026
- 0.8.7 14 Jul 2026
- 0.8.6 14 Jul 2026
- 0.8.5 14 Jul 2026
- 0.8.4 13 Jul 2026
- 0.8.3 13 Jul 2026
- 0.8.2 13 Jul 2026
- 0.8.1 12 Jul 2026
- 0.8.0 12 Jul 2026
- 0.7.1 12 Jul 2026
- 0.7.0 12 Jul 2026
- 0.6.0 11 Jul 2026
- 0.5.2 11 Jul 2026
- 0.5.1 10 Jul 2026
- 0.5.0 10 Jul 2026
- 0.4.1
- 0.4.0
- 0.3.0
- 0.2.0 3 Jun 2026
- 0.1.0