Bug

openclaw doctor warns about missing local model file even when local memory search is working correctly with the default auto-resolved model.

Steps to reproduce

1. Configure memory search with provider: "local" and no explicit local.modelPath
2. Run openclaw doctor
3. Doctor warns: "Memory search provider is set to "local" but no local model file was found."
4. But openclaw memory status --deep shows everything working:
   • Provider: local
   • Model: hf:ggml-org/embeddinggemma-300m-qat-q8_0-GGUF/embeddinggemma-300m-qat-Q8_0.gguf
   • Embeddings: ready
   • 48/48 files indexed, 212 chunks

Root cause

In the doctor check, hasLocalEmbeddings(resolved.local) only inspects local.modelPath from config:

function hasLocalEmbeddings(local) {
  const modelPath = local.modelPath?.trim();
  if (!modelPath) return false;
  // ...
}

When local is {} (no explicit modelPath), it returns false immediately. But the runtime auto-resolves a default HuggingFace model (embeddinggemma-300m) and works fine.

Expected behavior

Doctor should either:

• Check the runtime-resolved model (same logic the memory search system uses), or
• Skip the warning when the default model would be auto-resolved

Workaround

Explicitly set the model path in config:

openclaw config set agents.defaults.memorySearch.local.modelPath "hf:ggml-org/embeddinggemma-300m-qat-q8_0-GGUF/embeddinggemma-300m-qat-Q8_0.gguf"

Environment

• OpenClaw 2026.3.1 (2a8ac97)
• macOS (Apple Silicon)
• node-llama-cpp with Metal GPU