detect_mode_availability accepts a pre-computed dep_status so the agent
probes once; interface and fallback paths now have content-level tests.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
utils/capabilities.py now owns interface detection and mode
availability; the agent delegates via detect_interfaces() and
detect_mode_availability(). The agent keeps config gating and
tool_details population to preserve its result shape exactly.
The moved fallback path uses utils.dependencies.check_tool instead of
the agent's old shutil.which fallback; check_tool also searches
Homebrew paths, a strict superset (strictly better detection).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Both silently fell back to static bundled TLEs after the removal of
routes.satellite._tle_cache.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- busy_timeout so concurrent app+agent writers wait instead of raising
- seed from _connect() so update-before-first-read can't drop the seed
- regression tests: seed ordering, concurrent writer, default DB path
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- pyproject.toml: sync missing deps (flask-wtf, flask-compress,
simple-websocket, gunicorn, gevent, psutil, cryptography, meshcore,
pre-commit) so test_requirements integrity check passes
- tests/conftest.py: set INTERCEPT_DISABLE_AUTH=1 so auth routes
return 200 instead of 302 in tests
- routes/bluetooth_v2.py: add device_to_dict() helper that flattens
heuristics to top level for test_bluetooth_api serialization tests
- utils/bluetooth/heuristics.py: evaluate() now returns the device so
callers can chain; was returning None
- tests/test_satellite.py: reduce hours 48→2 in pass-prediction test
to prevent OOM kill on GitHub Actions 7GB runner at the 59% mark
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Move fingerprint stability update before early return so it updates even when payload hash matches
- Remove duplicate stability assignment from detect_tracker result block
- Add assertion in test to verify tracker fields are preserved when detection is skipped
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Modify DataStore.cleanup() to minimize lock hold duration:
- Snapshot timestamps under lock (brief O(1) list copy)
- Compute expired keys outside lock (no contention during O(n) scan)
- Re-acquire lock only for deletion with re-validation
(ensures entries refreshed between snapshot and deletion are not deleted)
This reduces blocking of reader threads and prevents latency spikes
during periodic cleanup of large stores (10K+ entries).
Also adds tests:
- test_cleanup_removes_expired_keeps_fresh: basic cleanup behavior
- test_cleanup_does_not_delete_refreshed_entry: re-validation guard
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Enable debug=True on MeshCore.create_ble() to surface verbose logs
- Disconnect any existing BlueZ connection before bleak connects to
avoid conflicts from prior bluetoothctl/pairing sessions
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Store last status message on MeshcoreClient so error details survive
beyond the SSE event (which isn't active during connecting state)
- Status endpoint now returns message field so the frontend can show
the real reason (e.g. 'Connection failed after retries: ...')
- Extend JS polling from 30s to 90s to outlast the backend's 65s
retry sequence (5+15+45s delays) before declaring timeout
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
asyncio.run() called from a gevent-patched Flask thread fails under
gunicorn+gevent. Run the one-shot scan in a ThreadPoolExecutor thread
with its own event loop, matching how AsyncWorker handles it.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Revert route to scan_ble() — scan_ble_sync() lives on AsyncWorker,
not MeshcoreClient; the 500 was caused by our previous fix
- MeshcoreClient.scan_ble() now runs a one-shot asyncio scan when no
worker is active, so Scan works before Connect is pressed
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Log node_id hint in request_traceroute instead of silently dropping it
- Replace asyncio.shield/wait_for pattern with _wait_or_stop() to prevent orphan tasks on retry delays
- Poll _stop_event every 1s in _do_connect keep-alive loop to handle stop() race before _asyncio_stop is set
- Extract pubkey_prefix/sender_id in _on_channel_msg instead of hardcoding "unknown"
- Close coroutine and log in _submit() when worker is not running to prevent ResourceWarning
- Cap battery_pct at 100 to prevent values exceeding 100%
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Implements AsyncWorker — the daemon asyncio thread that owns the meshcore
library connection, subscribes to all relevant EventTypes, and feeds events
back into MeshcoreClient via on_message/on_node/on_telemetry/on_traceroute/
on_connected/on_error. Includes retry-with-backoff (3 attempts: 5s/15s/45s),
thread-safe send_text/request_traceroute/scan_ble_sync for Flask callers,
and a standalone _scan_ble() coroutine using bleak.BleakScanner.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Lock-protect `get_state` and `_set_state` to prevent data race
between Flask and asyncio daemon threads
- Atomically check-and-set CONNECTING guard in `connect()` to close
TOCTOU window between concurrent Flask threads
- Push status events outside the lock in both `_set_state` and
`connect()` to avoid potential deadlock
- Add TestMeshcoreContact, TestMeshcoreClientStateMachine tests
covering to_dict keys, queue push on state change, message append
and 500-item cap (9 -> 13 tests)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Implements utils/meshcore.py with all dataclasses (MeshcoreMessage,
MeshcoreNode, MeshcoreContact, MeshcoreTelemetry, MeshcoreTraceroute),
connection configs (SerialConfig, TCPConfig, BLEConfig), ConnectionState
enum, serial port discovery, and the MeshcoreClient singleton skeleton.
Adds tests/test_meshcore_client.py covering all dataclasses, availability
check, and state enum (8/8 tests passing).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two root causes behind HackRF showing as unavailable when tools are installed:
1. get_tool_path() didn't search /usr/local/bin on Linux. HackRF tools built
from source (as in the Dockerfile) land there, but the path wasn't checked
when sudo/service environments have a restricted PATH.
2. check_hackrf() only tested hackrf_transfer, but the health check tests
hackrf_info — both come from the same apt package but a user could have one
visible and not the other. Now either binary confirms the tools are present.
hackrf_transfer is still required for actual RX/TX operations.
Fixes#212
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Data pipeline (critical): scanners/detectors now write to a separate _obs_queue;
a relay thread reads observations and calls correlator.process(), which emits
processed DroneContact dicts to drone_queue for SSE. Without this the SSE stream
received raw unserializable dataclass objects causing JSON errors.
Frontend (critical):
- Add droneContactList container to drone.html so contact cards render
- Add droneMap container and initialize Leaflet in drone.js init()
- Define dsc-distress-pulse keyframes in drone.css (was referenced but missing)
- Fix SSE reconnect: null _sse before setTimeout to prevent _connectSSE no-op loop
Other fixes:
- Validate rtl_sdr_index with validate_device_index(), return 400 on bad input
- Move _ensure_workers() inside _drone_lock to prevent double-initialization race
- Add double-call guard to RemoteIDScanner.start()
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Replace _running bool with threading.Event for correct cross-thread visibility
- Add _proc_lock to guard _rtl_proc/_hackrf_proc across worker/main threads
- Use register_process + safe_terminate (pipe close + SIGKILL fallback on timeout)
- Compute HackRF frequency as band midpoint (hz_low+hz_high)//2, not hz_low
- Guard start() for idempotency — double-call no longer leaks threads
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Implements RFDetector class that wraps rtl_433 (433/868MHz) and hackrf_sweep
(2.4/5.8GHz) subprocesses, emitting RFObservation objects onto a shared queue.
Includes signature matching, frequency band validation, and power thresholding.
- _handle_rtl433_line(): Parse JSON output, filter drone bands, emit observations
- _handle_hackrf_line(): Parse CSV output, average power levels, threshold at -90dBm
- start()/stop(): Manage subprocess threads for concurrent RF detection
- Graceful handling of missing tools (rtl_433, hackrf_sweep)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Prevents a non-running sniffer object being stored when start() raises
(e.g. permission denied or interface not found).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two root causes behind HackRF showing as unavailable when tools are installed:
1. get_tool_path() didn't search /usr/local/bin on Linux. HackRF tools built
from source (as in the Dockerfile) land there, but the path wasn't checked
when sudo/service environments have a restricted PATH.
2. check_hackrf() only tested hackrf_transfer, but the health check tests
hackrf_info — both come from the same apt package but a user could have one
visible and not the other. Now either binary confirms the tools are present.
hackrf_transfer is still required for actual RX/TX operations.
Fixes#212
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Sync with upstream main and fix required items from review:
- updateTimelineLabels() now uses InterceptTime API (getTimezone/getIANA)
instead of the stale selectedTimezone/TZ_MAP globals that were removed
during the earlier InterceptTime refactor — fixes ReferenceError on TZ
change and pass refresh.
- Remove profiles: [basic] from the intercept service in
docker-compose.yml so bare `docker compose up -d` still starts the
main service. Profile-gated services (intercept-history, adsb_db)
stay as-is.
data/*.json was excluded by .dockerignore, so wefax_stations.json was
never copied into the container image. The volume mounts in docker-compose
only cover subdirectories (weather_sat, adsb, etc.), leaving the stations
file inaccessible at runtime — causing the /wefax/stations route to 500
and the station/frequency dropdowns to appear empty.
Also adds a graceful file-existence check in load_stations() so a missing
file logs a warning and returns an empty list instead of an unhandled
FileNotFoundError.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(adsb): disable bias-T on stop and warn when toggled while running
The RTL-SDR bias-T hardware register persists after the device is closed,
so toggling bias-T off in the UI and stopping the SDR had no effect on the
actual hardware — verified with a multimeter in issue #205.
- Add disable_bias_t_via_rtl_biast() to rtlsdr.py (mirrors enable, uses -b 0)
- Track adsb_bias_t_active in adsb.py; call disable on stop_adsb() so the
hardware register is cleared when ADS-B is stopped
- Show an inline warning in the UI when the bias-T checkbox is toggled while
any SDR mode is active, since the setting only takes effect at start time
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(lint): remove unused imports in tscm sweep.py
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
AIS:
- New optional NMEA UDP forwarding via AIS-catcher's -u flag, configurable
from the AIS sidebar (host + port). Lets OpenCPN and other NMEA tools
receive live vessel data directly. All SDR builders updated.
- New GET /ais/vessels endpoint — clean JSON snapshot of tracked vessels
for REST integration
ADS-B:
- New GET /adsb/aircraft endpoint — JSON snapshot of all tracked aircraft,
with optional ?icao= and ?military=true filters. Response includes a
reminder that port 30003 (SBS) is already available for tools like
Virtual Radar Server and OpenCPN's AIS/target plugin.
Closes#90
- Pager and sensor gain inputs changed from unvalidated text fields to
number inputs with min/max/step constraints
- ADS-B dashboard now exposes a gain input in the tracking strip;
previously gain was hardcoded to 40 dB with no user control
- validate_gain() ceiling raised from 50 to 102 dB to support HackRF
(LNA 40 + VGA 62 = 102 dB combined) and LimeSDR (73 dB)
- sdrCapabilities gain_max values corrected: HackRF 62→102, Airspy 21→45
- onSDRTypeChanged() now propagates gain_max to all mode gain inputs so
HTML constraints match the selected SDR's actual range
Closes#162
Docker fixes:
- Add missing COPY for /usr/local/share/ (pipeline definitions were never
reaching the runtime image — root cause of silent SatDump failures)
- Add libfftw3-double3 and libfftw3-single3 runtime dependencies
- Handle arm64 vs x86 install path differences (/usr vs /usr/local)
- Split SatDump compile and staging into separate layers for better caching
- Add build-time assertions to catch missing pipelines early
UI enhancements:
- Timezone selector (UTC, Local, Eastern, Central, Mountain, Pacific)
with localStorage persistence — all time displays update instantly
- Pass analysis bar showing 24h quality breakdown and best upcoming pass
- Enhanced pass cards with cardinal direction (NW→SE), BEST badge
- Console timestamps, log level filters (ALL/SIGNAL/PROG/ERR), COPY/CLR
- Pass count in stats strip
- Demo data mode for UI testing without SDR or live satellite pass
- Meteor M2-4 80k baud fallback pipeline option
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move aircraft_db.json and aircraft_db_meta.json from the project root
to data/adsb/ so they survive container restarts and rebuilds. Add
matching volume mount to both Docker Compose profiles.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Fix deep scan with 'All bands' never scanning 5GHz: band='all' now
correctly passes --band abg to airodump-ng (previously no flag was
added, causing airodump-ng to default to 2.4GHz-only)
- Fix APs first seen without channel info permanently stuck at
band='unknown': _update_access_point now backfills channel, frequency,
and band when a subsequent observation resolves the channel
- Fix legacy /wifi/scan/start combining mutually exclusive --band and -c
flags: --band is now only added when no explicit channel list is given,
and the interface is always placed as the last argument
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
James Smith