- Replace send_message_tool import with hermes send CLI subprocess
- Remove broken _ensure_gw_config() that clobbered env vars
- Remove unused CircuitBreaker class and tenacity import
- Clean up docstring and unused imports
- Token fix: requires real TELEGRAM_BOT_TOKEN in ~/.hermes/.env
- Systemd drop-in: /etc/systemd/system/hermes-relay.service.d/telegram.conf
- Migrate from Flask to Quart (async) for better concurrency
- Add httpx with HTTP/2 connection pooling for Telegram
- Add diskcache for caching health check responses
- Increase ThreadPoolExecutor workers from 4 to 8 (configurable)
- Remove hardcoded paths - use environment variables
- Add circuit breaker pattern for resilient external calls
- Add proper timeout handling for health checks
- Pin all dependencies in requirements.txt
- Add graceful startup/shutdown handlers
- Pre-warm agent in background thread on startup
- Use voice-assistant profile (minimal chatbot, max_turns=10, nous provider)
- Remove /fast prefix and model override - handled by profile config
- Use messaging toolset for send_message tool (Telegram delivery)
- Add hardcoded TELEGRAM_BOT_TOKEN and TELEGRAM_HOME_CHANNEL to relay env
- Telegram delivery now works via gateway's send_message_tool
- Version bump to 2.1
- Health endpoint shows profile: voice-assistant
- Vervang subprocess calls naar hermes chat -q door directe Python import
- Nieuwe hermes_client.py met warme AIAgent (eenmalig init, hergebruik per query)
- Initializing agent overhead: 3-5s per query -> 0.48s eenmalig
- Queries na eerste zijn ~5s (alleen LLM API tijd)
- Auxiliary vision provider leeggezet (geen DNS lookup naar Nous)
wouser