The before_request require_login hook was returning a 302 redirect
for WebSocket upgrade requests, which browsers report as "Invalid
frame header". WebSocket requests don't always carry session cookies
reliably. Allow /ws/ paths through the login check since the page
that initiates these connections already requires authentication.
Also keeps the prior fix: serialize WebSocket sends through a queue
to avoid concurrent read/write on the non-thread-safe simple-websocket.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The fft_reader thread was calling ws.send() concurrently with
ws.receive() in the main loop. simple-websocket is not thread-safe
for simultaneous read/write, corrupting frame headers. Now the reader
thread enqueues frames and only the main loop touches the WebSocket.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace the batch rtl_power SSE pipeline with continuous I/Q streaming
via WebSocket for smooth ~25fps waterfall display. The server captures
raw I/Q samples (rtl_sdr/rx_sdr), computes Hann-windowed FFT, and
sends compact binary frames (1035 bytes vs ~15KB JSON, 93% reduction).
Client falls back to existing SSE path if WebSocket is unavailable.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Waterfall clicks now auto-select the correct modulation for the frequency
band (e.g., WFM for FM broadcast, AM for airband) instead of using whatever
modulation was last selected. Adds a hover tooltip showing frequency and
suggested modulation. Fixes the kill-all notification to show a clean
"All processes stopped" message instead of listing "bluetooth_scanner".
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds a phosphor-persistence waveform scope showing audio RMS/peak
levels during ISS SSTV and General SSTV decoding, matching the
existing pager scope pattern with a purple color scheme.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Enable -M level on rtl_433 to include RSSI/SNR in decoded JSON, extract
signal levels and push scope events to the SSE stream. Renders a green-
themed canvas oscilloscope showing signal strength pulses on packet decode
with amber SNR indicator and decay between packets.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Tap the rtl_fm → multimon-ng audio pipeline via a relay thread to extract
RMS/peak amplitude levels and render a 60fps canvas oscilloscope during
pager decoding, giving visual feedback of RF activity before messages are
fully decoded.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Reverts IQ pipeline and removes syncWaterfallToFrequency calls from
pager, sensor, rtlamr, DMR, SSTV, and SSTV general modes. Waterfall
is now exclusive to listening post mode.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace rtl_fm/rtl_433 with rtl_sdr for raw IQ capture when available,
enabling a Python IQ processor to compute FFT for the waterfall while
simultaneously feeding decoded data to multimon-ng (pager) or rtl_433
(sensor). Falls back to the legacy pipeline when rtl_sdr is unavailable.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Kill audio pipeline when startup produces no data instead of leaving
zombie processes running
- Skip unnecessary 1s USB release delay when no processes were active
- Remove racy fresh=1 pipeline restart from stream endpoint
- Await stopScanner() before starting direct listen to prevent race
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Allow users to override the pgdata volume mount via PGDATA_PATH env var,
enabling external storage (e.g. USB) for ADS-B history. Defaults to
./pgdata for backwards compatibility.
Based on PR #88 by JamesIOmete, rebased cleanly onto main.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Stream threads for sensor, pager, acars, rtlamr, dmr, and dsc modes
never called release_sdr_device() when their SDR process crashed,
leaving devices permanently locked in the registry. Also fixes orphaned
companion processes (rtl_fm, rtl_tcp) not being killed on crash, start
path failures leaking processes, DMR stop handler missing lock, and
listening post/audio websocket pkill nuking all system-wide rtl_fm
processes. Wires up register_process()/unregister_process() so the
atexit/signal cleanup safety net actually works, and adds rtl_tcp,
rtl_power, rtlamr, ffmpeg to the killall endpoint.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The RiskLevel.NEEDS_REVIEW enum value was 'review' but the
devices_by_risk dict and all summary keys used 'needs_review',
causing a KeyError during sweep correlation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fix infinite CPU spin in PD120 decoding caused by a 1-sample rounding
mismatch between line_samples (24407) and the sum of sub-component
samples (24408). The feed() while loop would re-enter _decode_line()
endlessly when the buffer was too short by 1 sample. Added a stall
guard that breaks the loop when no progress is made.
Fix false "leader tone detected" in the signal monitor by requiring
the detected tone to dominate the other tone by 2x, matching the
approach already used by the VIS detector.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When the WiFi interface is down (e.g. USB adapter not activated),
scanning fails with "Network is down" errors. Now the scanner
proactively checks interface state via /sys/class/net and brings
it up using ip link (or ifconfig fallback) before attempting scans,
with a retry loop if the initial scan still fails.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The Receiver Count section had no <h3> so it didn't get collapsible
panel styling, rendering as a small out-of-place rectangle. The count
is already shown in the main receiver list panel so this was redundant.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Static minZoom: 2 wasn't enough for tall containers. Now calculate
minZoom from actual container height so tiles always cover the visible
area. Also set map background to match CartoDB dark tile ocean color
so any remaining edge at extreme latitudes blends seamlessly.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add maxBounds to limit vertical panning to ±85° latitude and set
minZoom to 2 so tiles always cover the visible area. Prevents the
large black bands above and below the map tiles.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The dmrRawOutput div was rendering garbled box-drawing characters from
the dsd-fme ASCII art banner below the signal activity canvas. Remove
the div and filter banner lines (box-drawing chars, version info) in
the parser so they never become events.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The stream thread used a blocking readline() with no timeout, so once
DSD finished outputting its startup banner there were no more events
until actual signal activity. The frontend decayed to zero and appeared
dead. If DSD crashed, the synthesizer state never transitioned to
'stopped' so there was no visual or textual indication of failure.
- Use select() with 1s timeout on DSD stderr to avoid indefinite block
- Send heartbeat events every 3s while decoder is alive but idle
- Detect DSD crashes: capture exit code and remaining stderr, send as
'crashed' status with details and show notification to user
- Frontend properly transitions synthesizer to 'stopped' on process
death (was only happening on user-initiated stop)
- Increase idle breathing amplitude so LISTENING state is clearly
visible (0.12 +/- 0.06 vs old 0.05 +/- 0.035)
- Release device reservation on crash, not just user stop
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
DMR was missing checkDeviceAvailability/reserveDevice/releaseDevice
calls that other modes (SSTV, listening post) use, so the device
dropdown showed device 0 as available even when another process held
it. Also detect USB claim errors from rtl_fm and surface a clear
message telling the user to pick a different device.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The DSD stderr parser had regex ordering bugs that swallowed voice and
call events as bare slot events, and only matched classic dsd output
format (not dsd-fme). Unmatched lines were silently dropped, leaving
the signal activity panel with nothing to display.
- Reorder regex checks: TG/Src before voice before slot
- Support dsd-fme comma-separated format (TG: x, Src: y)
- Make bare slot regex strict (only standalone "Slot N" lines)
- Forward unmatched DSD lines as raw events for diagnostics
- Add LISTENING state to signal activity panel for raw output
- Show raw decoder output text below synthesizer canvas
- Fix test mocks for find_dsd() tuple return value
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The APRS route called app_module.reserve_sdr_device() which does not
exist, causing an AttributeError that Flask returned as an HTML error
page. The frontend then failed to parse it as JSON, showing
"Unexpected token '<'" to the user. Fixed to use claim_sdr_device()
which is the correct function used by all other modes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
SSTV was not claiming/releasing SDR devices through the centralized
registry, so the device state panel always showed the device as idle
during SSTV use. Added claim_sdr_device/release_sdr_device on the
backend and reserveDevice/releaseDevice on the frontend, matching the
pattern used by all other modes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Users can now manage decoded SSTV images with download and delete actions
accessible from hover overlays on gallery cards, the full-size image modal
toolbar, and a "Clear All" button in the gallery header. Both ISS and
General SSTV modes are supported.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The decode canvas was always black because nothing drew on it. Now the
backend encodes partial JPEG snapshots every 5% progress and the frontend
uses an <img> tag with in-place DOM updates instead of recreating innerHTML
on every SSE event.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Shows the current VIS detection state machine position (Idle, Leader,
Break, Start bit, Data bits, etc.) in the signal monitor. This helps
diagnose why decoding may not be starting - e.g. if the VIS detector
is stuck in Idle despite a leader tone being present, the signal may
not contain a valid VIS header.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The Doppler tracking thread emits detecting events every 5s from a
separate thread, unaware of decode state. The previous to_dict() change
included signal_level for ALL detecting events, causing the frontend to
replace the decode progress canvas with the signal monitor mid-decode.
Fix: use None as default for signal_level so only signal-metrics events
(which explicitly set the value) include the field. Also add a frontend
guard to ignore detecting events while the UI is in decoding state.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The to_dict() method was skipping signal_level when it was 0, so the
frontend never received the field and never rendered the monitor.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Shows RMS audio level bar and SSTV tone classification (leader/sync/noise)
via SSE during detecting mode, replacing the static "Listening..." state
with actionable signal feedback.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When rtl_fm exits unexpectedly, read its stderr output to diagnose
the failure (no device, permission denied, etc.) and include the
error message in both the server log and the SSE progress event
sent to the browser.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Three bugs preventing the live SSTV pipeline from working:
1. Race condition: self._running was set AFTER starting the decode
thread, so the thread checked the flag, found it False, and exited
immediately without ever processing audio.
2. Ghost running state: when the decode thread exited (e.g. rtl_fm
died), self._running stayed True. The decoder reported as running
but was dead, and subsequent start() calls returned without doing
anything - permanently stuck until app restart.
3. VIS detection fragility: unclassifiable windows at tone transition
boundaries (mixed energy from two tones) caused the state machine
to reset from LEADER/BREAK states back to IDLE, dropping valid
VIS headers on real signals.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
slowrx is a GTK GUI app that doesn't support CLI usage, so the SSTV
decoder was silently failing. This replaces it with a pure Python
implementation using numpy and Pillow that supports Robot36/72,
Martin1/2, Scottie1/2, and PD120/180 modes via VIS header auto-detection.
Key implementation details:
- Generalized Goertzel (DTFT) for exact-frequency tone detection
- Vectorized batch Goertzel for real-time pixel decoding performance
- Overlapping analysis windows for short-window frequency estimation
- VIS header detection state machine with parity validation
- Per-line sync re-synchronization for drift tolerance
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
getBoundingClientRect on the canvas itself (sized via CSS width:100%)
instead of parentElement with arbitrary offset, preventing zero-width
canvas when flex layout timing varies.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Event-driven spring-physics bar visualization reacting to SSE events
(sync/call/voice) with HSL color coding and center-outward ripple effects.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
rtl_fm stderr was sent to DEVNULL, hiding the actual failure reason
(rc=1). Now captured and surfaced in the error response. Also drains
rtl_fm stderr during normal operation to prevent pipe blocking.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Use -o - (stdout) instead of -o /dev/null for audio output, as
dsd-fme expects specific output targets. Remove -N flag which may
cause issues in headless mode. Add stderr capture on pipeline
failure for better error messages.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
dsd-fme uses different protocol flags than classic dsd (e.g. -fs for
DMR instead of -fd, -f1 for P25 instead of -fp). Add -N flag to
disable ncurses terminal which is required when reading from stdin pipe.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
