#Memory Search Not Working
1 messages · Page 1 of 1 (latest)
node-llama-cpp being “missing” is expected with the newer releases if you’re using local memory embeddings: it was changed to an optional peer dependency, and (especially on Windows) OpenClaw no longer tries to auto-install it during a normal npm i -g openclaw@latest update.
1) First check what provider your memory search is trying to use
Run:
openclaw config get agents.defaults.memorySearch.provider
openclaw config get agents.defaults.memorySearch.local.modelPath
If provider is local (or you have a local.modelPath set), then you need node-llama-cpp installed.
2) Fix options
Option A (simplest): switch to remote embeddings (no node-llama-cpp)
Set a provider like OpenAI/Gemini and add the API key:
openclaw config set agents.defaults.memorySearch.provider "openai"
# put OPENAI_API_KEY=... into ~/.openclaw/.env (or your service env)
openclaw gateway restart
Docs: https://docs.openclaw.ai/concepts/memory
Option B: keep local embeddings (install node-llama-cpp)
-
If you installed with npm:
npm i -g node-llama-cpp@3.16.2 openclaw gateway restart -
If you installed with pnpm: