Excessive log volume: ~900MB/day from observation permalink resolution attempts

[bm-clawd]

Bug

The BM MCP server generates ~9.5MB of logs every 7 minutes (~900MB/day, 1GB currently on disk) at INFO level during normal usage with the OpenClaw plugin.

Root Cause

The composited memory_search in the openclaw-basic-memory plugin queries active tasks, which triggers resolution of every observation permalink on every task. Each observation path like claw/memory/tasks/fix-graph-visualization-pr-469/observations/started/2026-02-24 gets a resolve_identifier call that fails (these aren't real entities), generating multiple INFO log lines per attempt.

With ~10 active tasks × ~10 observations each × multiple search calls per heartbeat (every 7 min), this produces thousands of failed lookups per cycle.

Evidence

In a single 9.5MB log file (covering ~7 minutes):

• 13,462 lines mentioning openclaw-basics-mbp-lan
• 8,504 get_client calls
• 168 identical failed lookups for each task observation path

Impact

• 100 rotated log files in ~/.basic-memory/ (1GB total)
• retention="10 days" means this could grow to 9GB
• Disk pressure on machines with limited storage

Suggestions

1. Reduce log level for failed resolve_identifier lookups from INFO to DEBUG
2. Add retention file count limit in addition to time-based retention (e.g., retention=10 files max)
3. The observation permalink resolution is the real fix — the plugin shouldn't be trying to resolve observation sub-paths as entities

Environment

• basic-memory 0.18.5
• openclaw-basic-memory 0.1.0-alpha.12 (but log issue predates this — bm watch from Feb 18 was also contributing)
• Log config: rotation="10 MB", retention="10 days", level=INFO

[claude]

Claude encountered an error —— View job

---

I'll analyze this and get back to you.