Summary

• Fixes false positive warning in openclaw doctor when using QMD memory backend with local embeddings (e.g. embeddinggemma)
• Adds an early return in noteMemorySearchHealth() when memory.backend === "qmd"

Fixes #17263

Problem

When QMD is configured as the memory backend, it handles embeddings internally — no separate embedding provider needs to be configured in OpenClaw's memorySearch settings. However, openclaw doctor doesn't know this and falsely warns:

Memory search is enabled but no embedding provider is configured.

Root cause

noteMemorySearchHealth() checks hasLocalEmbeddings(resolved.local) which looks for local.modelPath. QMD doesn't use this field — it bundles its own embedding model. The function then checks for remote API keys (OpenAI/Gemini/Voyage) and finds none, triggering the false positive.

Fix

Call resolveMemoryBackendConfig() at the top of the check. If backend === "qmd", return early — QMD handles its own embeddings, so no warning is needed.

1 file changed: src/commands/doctor-memory-search.ts

Test plan

• Configure QMD as memory backend with local embeddings
• Run openclaw doctor — verify no false warning about missing embedding provider
• Configure builtin backend without embedding provider — verify warning still appears
• Configure builtin backend with OpenAI key — verify no warning

Local Validation

• pnpm check (tsgo): ✅ passes
• pnpm test: ✅ passes (existing doctor tests still green)
• Formatting: verified with oxfmt --check

Scope

Single-file fix: src/commands/doctor-memory-search.ts — adds early return when QMD backend is detected.

AI Assistance

AI-assisted (Claude Code) for codebase exploration and root cause analysis. The diagnostic (QMD bundles its own embeddings, so the provider check is unnecessary) and the fix approach are my own work.

Testing level: Fully tested — confirmed pnpm check and pnpm test pass locally.

Author

Miloud Belarebia — @miloudbelarebia

Greptile Summary

Fixes a false positive warning in openclaw doctor when using the QMD memory backend. QMD bundles its own embedding model (e.g. embeddinggemma), so the existing embedding provider check in noteMemorySearchHealth() incorrectly flagged a missing provider. The fix adds an early return when resolveMemoryBackendConfig() reports backend === "qmd", skipping the irrelevant provider check entirely.

• Adds resolveMemoryBackendConfig call in noteMemorySearchHealth() with early return for QMD backend
• Adds properly hoisted mock and test case for the QMD backend path, addressing both items from prior review feedback
• Existing tests continue to pass with a default { backend: "builtin" } mock in beforeEach

Confidence Score: 5/5

• This PR is safe to merge — it adds a narrow, well-tested early return that only affects QMD backend users.
• The change is minimal (one new import, one early-return guard) with clear intent. The logic is correct: MemoryBackend is a union of "builtin" | "qmd", and QMD handles its own embeddings as confirmed by the backend-config.ts source and usage in other callers like search-manager.ts. The test file properly mocks the new dependency and adds a specific test case. No regressions to existing tests since the default mock returns { backend: "builtin" }. No security concerns.
• No files require special attention

_{Last reviewed commit: d970271}