Imported from soerensen3/neverending-quarrel (skills/redis-logger/SKILL.md). Install upstream with npx skills add soerensen3/neverending-quarrel --skill redis-logger. Copyright stays with the author.

redis-logger

After generating every response, mark speaking state and log output to local Redis.

Required workflow

1. Read MACHINE_ID from environment.
2. Read REDIS_HOST from environment.
3. Set local speaking flag before TTS starts:

redis-cli -h "$REDIS_HOST" SET agent:speaking 1 EX 30

4. Build a JSON payload with:
   • source: value of MACHINE_ID
   • text: exact response text that was generated
   • timestamp: current Unix timestamp in milliseconds
   • kind: assistant
5. Push the payload to Redis:

redis-cli -h "$REDIS_HOST" LPUSH conversation:log "$JSON_PAYLOAD"

6. Store the latest assistant text for echo suppression helpers:

redis-cli -h "$REDIS_HOST" SET agent:last_tts_text "$RESPONSE_TEXT" EX 120
redis-cli -h "$REDIS_HOST" SET agent:last_tts_ts_ms "$NOW_MS" EX 120

7. Trim the list so only the latest 100 entries remain:

redis-cli -h "$REDIS_HOST" LTRIM conversation:log 0 99

8. Clear speaking flag after playback finishes and apply a short cooldown:

redis-cli -h "$REDIS_HOST" SET agent:speaking 0 EX 2
redis-cli -h "$REDIS_HOST" SET agent:cooldown_until_ms "$((NOW_MS + 1200))" EX 5

Notes:

• This Redis is local to each machine. Do not assume cross-machine connectivity.
• If TTS crashes mid-playback, let the agent:speaking TTL expire naturally.