1 Commits (develop)

Author	SHA1	Message	Date
Joaquin Madrid Belando	aa96b48359	Mejoras de Audio y funconalidad en caliente voice.cfg (#27) ... * Update adn.cfg * Update replit.md * Update config.py * Update bridge_master.py * Fix: corregir indentacion en llamada a _handleRecording linea 2846 * Mover configuracion de locuciones y grabaciones a voice.cfg * Nuevo archivo voice.cfg con configuracion de locuciones y grabaciones * Cargar locuciones/grabaciones desde voice.cfg (opcional, con fallback a defaults) * Actualizar documentacion: voice.cfg separado de adn.cfg * Agregar configuracion TTS (4 slots) en voice.cfg * Leer configuracion TTS desde voice.cfg * Nuevo modulo TTS: texto -> gTTS -> WAV -> vocoder -> AMBE * Agregar gTTS a dependencias * Integrar sistema TTS con scheduling y playback * Agregar configuracion AMBEServer (DV3000 remoto via UDP) en voice.cfg * Leer configuracion TTS_AMBESERVER_HOST/PORT desde voice.cfg * Agregar cliente AMBEServer UDP (protocolo DV3000) para codificacion AMBE * Fix: apply DMR 3-way interleave to AMBE frames from DV3000 DV3000 outputs raw 72-bit AMBE+2 frames but DMR voice bursts require 3 frames interleaved into 216-bit payloads. Interleave: burst_payload[i*3+j] = frame[j].bit[i] (ETSI TS 102 361-1) Also uses proper AMBE silence frame for padding incomplete triplets. * Fix: use RATET 33 (DMR AMBE+2) instead of RATEP (D-Star) The DV3000 was being configured with RATEP parameters that correspond to D-Star AMBE codec, not DMR AMBE+2. G4KLX AMBETools uses RATET 33 (rate table entry 33) for DMR which selects the correct codec. Also removed incorrect interleaving - DMR voice bursts use raw AMBE frames without additional interleaving (confirmed by G4KLX wav2ambe). * Add reload_voice_config() for hot-reload of voice.cfg * Hot-reload voice.cfg: auto-detect changes every 15s without restart Checks voice.cfg mtime every 15 seconds. When changed, reloads all voice/TTS config, stops old LoopingCall tasks and starts new ones with updated settings. No server restart needed. * Fix: resolve AMBEServer host explicitly, strip whitespace/quotes Adds socket.gethostbyname() to resolve DNS or validate IP before using it in sendto(). Also strips whitespace and quotes from host value to prevent hidden characters from config parsing. * Change default TG from 214 to 2 in voice.cfg * Improve playFileOnRequest: add file existence check, better logging, catch all exceptions * Filter announcements by TG: only send to peers active on matching talkgroup * Filter announcements/TTS by active BRIDGES: only send to systems with static or active dynamic TG * Add TTS_VOLUME parameter (tts_engine.py): adjustable dB volume for TTS audio * Add TTS_VOLUME parameter (config.py): adjustable dB volume for TTS audio * Add TTS_VOLUME parameter (config/voice.cfg): adjustable dB volume for TTS audio * Run TTS conversion (gTTS+ffmpeg+AMBEServer) in separate thread via deferToThread to avoid blocking reactor * TTS dual-slot: send announcements on both TS1 and TS2 based on active BRIDGES per system * Remove TTS_ANNOUNCEMENT*_TIMESLOT: TS now auto-detected from BRIDGES (config.py) * Remove TTS_ANNOUNCEMENT*_TIMESLOT: TS now auto-detected from BRIDGES (config/voice.cfg) * Remove TTS_ANNOUNCEMENT*_TIMESLOT: TS now auto-detected from BRIDGES (bridge_master.py) * Announcements dual-slot: auto-detect TS from BRIDGES, remove ANNOUNCEMENT*_TIMESLOT (bridge_master.py) * Announcements dual-slot: auto-detect TS from BRIDGES, remove ANNOUNCEMENT*_TIMESLOT (config.py) * Announcements dual-slot: auto-detect TS from BRIDGES, remove ANNOUNCEMENT*_TIMESLOT (config/voice.cfg) * Fix on-demand playback timing drift: use absolute timing instead of sleep(0.058) to prevent audio cuts on long files * Add broadcast queue: prevent simultaneous announcements causing audio micro-cuts. All announcements (AMBE + TTS) now queue sequentially with 1.5s gap between them. * Fix frame interval: 54ms->60ms (DMR standard) + absolute timing on all voice functions to eliminate micro-cuts * Adjust frame interval: 60ms->58ms for optimal playback speed on all voice functions * Remove noisy ROUTER debug logs that block reactor and cause voice micro-cuts: eliminate hundreds of 'no change'/'NO ACTION' debug lines per cycle * Move rule_timer_loop, statTrimmer, kaReporting to background threads to prevent reactor blocking during voice broadcasts - all logs preserved * Fix RuntimeError: dictionary changed size during iteration - use list(BRIDGES) for thread-safe iteration in all functions that access BRIDGES * Revert bridge_master.py to state of commit ce9e9b09e0 * Skip rule_timer_loop during active voice traffic to prevent GIL contention micro-cuts - all logs preserved * Batch ROUTER debug logs into single write to reduce GIL contention during voice traffic - all log content preserved * Add TTS_SPEED config for speech rate control (atempo filter in ffmpeg) * Add TTS_SPEED config for speech rate control (atempo filter in ffmpeg) * Add TTS_SPEED config for speech rate control (atempo filter in ffmpeg)	2 weeks ago