memory-lancedb-pro appears functional, but openclaw status/doctor reports memory unavailable

[lamkan0210]

Summary

After upgrading OpenClaw to 2026.4.2, memory-lancedb-pro appears to be loaded and functioning, but openclaw status / openclaw doctor still report memory as unavailable or inactive.

This looks like a health-probe / active-plugin-detection mismatch rather than a hard runtime failure.

Environment

• OpenClaw: 2026.4.2
• OS: macOS 26.5 (arm64)
• Node: 25.6.0
• Memory plugin log shows: memory-lancedb-pro@1.1.0-beta.10

What I observed

Plugin startup logs show the plugin is being registered successfully:

[plugins] memory-lancedb-pro@1.1.0-beta.10: plugin registered (db: /Users/qianglin/.openclaw/memory/lancedb-pro-1024, model: bge-large:latest, smartExtraction: ON)
[plugins] memory-lancedb-pro: smart extraction enabled (LLM model: openai/gpt-oss-120b, timeoutMs: 30000, noise bank: ON)
[plugins] session-memory: hook registered for command:new as memory-lancedb-pro-session-memory

But openclaw status reports:

Memory: enabled (plugin memory-lancedb-pro) · unavailable

And openclaw doctor --non-interactive reports:

No active memory plugin is registered for the current config.

Why this seems inconsistent

From the logs, the plugin appears to be:

• loaded
• registered
• hooked into session memory
• configured with a valid DB + embedding model

So the issue seems to be one of:

• active memory plugin detection
• health probe / status reporting
• doctor/status checking a different registration path than the runtime actually uses

Expected behavior

If the plugin is successfully registered and usable, openclaw status / doctor should reflect that accurately.

Actual behavior

Runtime logs suggest the plugin is active, but status/doctor still say memory is unavailable or inactive.

Extra note

This may be related to the distinction between:

• plugin loaded/registered at runtime
• "active memory plugin" detection used by CLI health/status commands

If useful, I can provide more logs from openclaw status / openclaw doctor output.

[AliceLJY]

Thanks for the report. This looks like the same status/doctor integration gap behind #434, and PR #471 is the right place to fix it.

I just reviewed that PR and asked for one change before merge: the runtime stub cannot return hard-coded success probes, or doctor will go green even when the plugin itself is unhealthy. Once that probe path reflects the plugin's real health, this issue should be resolved.