Description

When the gateway starts/restarts, it runs qmd update to sync the semantic search index. For installations with large indexes (8,000+ files across multiple collections), the command output exceeds the hardcoded 200,000 character buffer limit, causing the boot sync to fail silently every restart.

[memory] qmd boot update failed: Error: qmd update produced too much output (limit 200000 chars)

Environment

• OpenClaw: 2026.2.22-2
• QMD: latest (via bun)
• Index: 8,308 files, 34,850 vectors across 7 collections
• qmd update output: ~283K characters (exceeds 200K limit)

Collections indexed

• WhatsApp archive (markdown)
• Telegram archive (markdown)
• Signal archive (markdown)
• Workspace docs
• Google Drive shared folder
• Shared knowledge base
• Apple Notes export (~1,689 notes)

Current behavior

1. Every gateway restart triggers qmd update
2. Output exceeds 200K char limit → boot sync fails
3. Background 15-minute incremental sync eventually catches up (produces less output since only changes)
4. Semantic search has stale results for ~15 minutes after every restart

Expected behavior

Boot sync should succeed regardless of index size. Possible fixes:

1. Raise or remove the 200K output limit for boot update operations
2. Add a --quiet or --summary-only flag to qmd update that suppresses per-file output and only shows the summary line
3. Stream/discard stdout and only capture stderr for error detection (the gateway doesn't use the stdout content, it only needs to know success/failure)
4. Allow configuring the output limit in memory.qmd.update config section (e.g. maxOutputChars)

Workaround

Accept the 15-minute gap. The background memory.qmd.update.interval cycle produces incremental output well under the limit and keeps the index current. Only the initial boot sync fails.

Impact

Low-medium. Semantic search works correctly 99% of the time. Only affects the first ~15 minutes after any gateway restart.