Session Janitor

Edit config.json after setup to tune thresholds:

| Key | Default | Description | |-----|---------|-------------| | trimMaxKB | 250 | Trim transcripts larger than this. Trim always fires when exceeded — there is no minimum pair count. | | keepPairs | 10 | Number of recent user/assistant pairs to keep after trim. If fewer pairs exist than this, all are kept (no skip). | | keepFullPairs | 2 | Most recent N assistant turns that keep full toolResult bodies (older turns are collapsed to summary lines) | | minArchivePairs | 5 | Informational only — no longer blocks trim. Sessions exceeding trimMaxKB are always trimmed. | | trimFullThresholdPct | 50 | Two-stage aggressive reduction when trimmed output still exceeds this % of trimMaxKB: (1) strip all assistant turns (tool args + thinking removed); (2) if still over threshold, drop all toolResult entries entirely. Set to 100 to disable. | | archiveRetentionDays | 7 | Delete old archives after this many days (applies to .reset.*, .deleted.*, .pre-trim.*, .bak-*, .purged.*, .emergency-*) | | keepPreTrimFiles | 3 | Max .pre-trim.* files to keep per session (oldest deleted immediately, regardless of retention period) | | orphanGraceMinutes | 30 | Wait before archiving orphan transcripts | | staleSubagentHours | 24 | Prune subagent session entries older than this | | staleCronSessionHours | 24 | Prune cron session entries older than this | | llmExtraction.enabled | true | Use LLM to extract memories from trimmed content | | llmExtraction.model | openclaw | Model identifier to use for extraction calls | | llmExtraction.maxInputChars | 20000 | Max characters of archived content sent to LLM | | llmExtraction.timeoutSecs | 60 | LLM API call timeout in seconds | | llmExtraction.maxMemories | 15 | Max memories to accept from a single LLM extraction | | llmExtraction.minArchived | 3 | Minimum archived messages required before running LLM extraction | | llmExtraction.maxPerRun | 1 | Max LLM extractions per cron cycle (cost control) | | memCli.enabled | false | Store extracted memories via mem CLI (requires mem) | | cronSchedule | */15 * * * * | How often to run | | sidecar.enabled | true | Offload large toolResult entries to .toolcache/ files | | sidecar.minEntryBytes | 5120 | Minimum toolResult content size (bytes) to trigger offload | | extractOnTrim.enabled | false | Fire async LLM memory extraction after each watcher-triggered trim | | extractOnTrim.minArchivedPairs | 3 | Minimum archived message pairs required before running extraction | | extractOnTrim.scene | auto | Memory scene tag to use (auto lets the LLM infer from context) | | extractOnTrim.salience | 0.5 | Default salience for extracted memories (0.1–1.0) | | watchdog.enabled | false | Run hung-session detector after each janitor pass | | watchdog.staleMinutes | 5 | Session updatedAt age threshold to consider stuck | | watchdog.alertSlack | true | Send Slack DM when a stuck session is detected | | watchdog.slackTarget | — | Slack user ID or channel to notify | | watchdog.autoRestart | false | Auto-trigger safe-restart script on detection (use with caution) | | watchdog.restartScriptSlack | ~/bin/safe-slack-restart.sh | Safe restart script path for Slack gateway | | watchdog.restartScriptDiscord | ~/bin/safe-gateway-restart.sh | Safe restart script path for Discord gateway |