ollama-memory-embeddings
por vidarbrekke
Coding Agents & IDEsResumen
Configure OpenClaw memory search to use Ollama as the embeddings server (OpenAI-compatible /v1/embeddings) instead.
Descripción
Qué hace esta skill
Configures OpenClaw memory search to use Ollama as the embeddings server via its OpenAI-compatible /v1/embeddings endpoint, replacing the built-in node-llama-cpp local GGUF loader. Supports multiple embedding models, optional import of existing local GGUFs into Ollama, config drift enforcement, and a launchd watchdog for auto-healing.
Offloading embedding generation to Ollama avoids in-process GGUF loading overhead while keeping all data local, and the drift watchdog ensures config stays consistent across gateway restarts.
Cuándo usarla
- Switching OpenClaw memory search from local GGUF loading to Ollama embeddings
- Selecting a different embedding model without manually editing config
- Importing an existing local embedding GGUF into Ollama for reuse
- Running scheduled config drift checks to keep embedding settings consistent
- Rebuilding the memory index after changing embedding providers
Flujo de trabajo de ejemplo
User runs install.sh to switch OpenClaw memory search from built-in GGUF to Ollama
- 1
Verifies Ollama is installed and reachable at 127.0.0.1:11434
- 2
Prompts user to select an embedding model such as nomic-embed-text or embeddinggemma
- 3
Updates agents.defaults.memorySearch in OpenClaw config with provider, model, baseUrl, and apiKey fields
- 4
Optionally triggers a memory reindex if the embedding fingerprint changed
- 5
Calls the Ollama embeddings endpoint and validates the response to confirm setup
OpenClaw memory search uses Ollama for embedding vector generation, with config backed up and verified
Requisitos
- Ollama installed and running locally
Instalación
clawhub install vidarbrekke/ollama-memory-embeddingsEjecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.