Release v0.8.5

Fixed

• Replace blocking run_polling() with low-level async API (Application.initialize/start/updater.start_polling) to resolve polling hang where run_polling() blocks indefinitely and cannot be interrupted
• Add getMe()-based watchdog that probes Telegram API reachability every 60 seconds; after 5 minutes of consecutive failures, stops the updater and restarts polling in-process
• Detect unexpected polling termination and automatically restart without process restart
• Graceful shutdown between restart cycles ensures clean Application teardown

Release v0.8.4

Fixed

• Auto-restart Telegram polling after unexpected exit (e.g. SDK crash triggering graceful shutdown) instead of silently stopping message reception
• Retry transient SDK errors (SIGTERM, SIGKILL, ConnectionRefused) once with automatic reconnection in process_message
• NetworkError now retries indefinitely with application rebuild instead of giving up after fixed attempts
• Rapid crash protection: exits only after 5 consecutive polling failures within 30 seconds each

Release v0.8.3

Fixed

• Add network retry logic for connection resilience

Release v0.8.2

Fixed

• Add event loop watchdog that detects zombie state (asyncio loop closed but process alive) and force-exits, allowing start.sh auto-restart to recover
• Enable launchd KeepAlive so the service auto-restarts even if start.sh itself exits (e.g. rapid crash limit)

Changed

• Enhanced --status command to detect inactive bots via log mtime checking, reporting detailed diagnostics instead of a misleading "running" status

Release v0.8.1

Fixed

• Volcengine voice transcription now deletes the temporary TOS object after ASR completes, preventing staged voice files from accumulating over time
• TOS cleanup failures are isolated to logs and no longer affect user-facing transcription replies

Changed

• Extended TOS uploader API to return uploaded object metadata (object_key + signed URL) for explicit post-transcription cleanup
• Added tests covering TOS object deletion on both success and failure paths

Release v0.8.0

Added

• macOS voice reply mode with TTS support: bot automatically replies with voice when user sends voice messages, using macOS say command + ffmpeg conversion
• Smart voice delivery strategy based on response length (voice-only, text+voice, or text-only fallback)
• VOICE_REPLY_PERSONA config for selecting macOS TTS voice persona

Fixed

• Voice reply mode gracefully falls back to text on non-macOS platforms

Changed

• Updated README documentation (EN/ZH) with voice reply mode usage guide

Release v0.7.0

Added

• Volcengine ASR support for voice transcription as an alternative to OpenAI Whisper
• TOS (Tencent Object Storage) upload flow for Volcengine ASR integration

Changed

• Added Star History chart to README files

Release v0.6.3

Changed

• Renamed project from "Telegram Skill Bot" to "Claude Telegram Bot Bridge"
• Updated project name in README.md, README-zh.md, and start.sh
• Changed version display from "Bot version" to "Bridge version"
• Simplified update notification to non-interactive text prompt

Release v0.6.2

Added

• Auto-update check on startup with 1-hour cache to detect new releases
• Interactive upgrade prompt when update is available (upgrade now / skip)
• --upgrade command for one-click bot updates via git pull and dependency reinstall
• Version comparison logic to determine if update is needed
• Graceful handling of network failures during update check

Changed

• Updated README.md and README-zh.md with upgrade command documentation
• Added auto-update feature to Operations section in documentation

Release v0.6.1

Changed

• Simplified bot command descriptions for better user experience in Telegram command menu