openclaw - 💡(How to fix) Fix [Feature]: Support configurable embedding provider for memory semantic search

Summary

Body: Feature Request: Configurable embedding provider Problem memory_search currently fails if no embedding provider is configured: No API key found for provider "openai". Use Case Users running local GPU inference with Ollama (mxbai-embed-large) do not want to rely on cloud embedding services. The infrastructure is ready; only the configuration opening is missing. Suggested Solution Add a memory.embedding or memory.embeddingProvider config field that accepts a provider/model ID string: Example: memory: { backend: "builtin", embeddingProvider: "ollama/mxbai-embed-large:latest" } memory_search would then use the specified provider for vectorizing queries instead of hardcoding "openai". Environment OpenClaw version: 2026.5.27 Ollama running locally with embedding models available (mxbai-embed-large) Local GPU available for embedding inference

Problem to solve

Body: Feature Request: Configurable embedding provider Problem memory_search currently fails if no embedding provider is configured: No API key found for provider "openai". Use Case Users running local GPU inference with Ollama (mxbai-embed-large) do not want to rely on cloud embedding services. The infrastructure is ready; only the configuration opening is missing. Suggested Solution Add a memory.embedding or memory.embeddingProvider config field that accepts a provider/model ID string: Example: memory: { backend: "builtin", embeddingProvider: "ollama/mxbai-embed-large:latest" } memory_search would then use the specified provider for vectorizing queries instead of hardcoding "openai". Environment OpenClaw version: 2026.5.27 Ollama running locally with embedding models available (mxbai-embed-large) Local GPU available for embedding inference

Proposed solution

Body: Feature Request: Configurable embedding provider Problem memory_search currently fails if no embedding provider is configured: No API key found for provider "openai". Use Case Users running local GPU inference with Ollama (mxbai-embed-large) do not want to rely on cloud embedding services. The infrastructure is ready; only the configuration opening is missing. Suggested Solution Add a memory.embedding or memory.embeddingProvider config field that accepts a provider/model ID string: Example: memory: { backend: "builtin", embeddingProvider: "ollama/mxbai-embed-large:latest" } memory_search would then use the specified provider for vectorizing queries instead of hardcoding "openai". Environment OpenClaw version: 2026.5.27 Ollama running locally with embedding models available (mxbai-embed-large) Local GPU available for embedding inference

Alternatives considered

No response

Impact

Body: Feature Request: Configurable embedding provider Problem memory_search currently fails if no embedding provider is configured: No API key found for provider "openai". Use Case Users running local GPU inference with Ollama (mxbai-embed-large) do not want to rely on cloud embedding services. The infrastructure is ready; only the configuration opening is missing. Suggested Solution Add a memory.embedding or memory.embeddingProvider config field that accepts a provider/model ID string: Example: memory: { backend: "builtin", embeddingProvider: "ollama/mxbai-embed-large:latest" } memory_search would then use the specified provider for vectorizing queries instead of hardcoding "openai". Environment OpenClaw version: 2026.5.27 Ollama running locally with embedding models available (mxbai-embed-large) Local GPU available for embedding inference

Evidence/examples

No response

Additional information

No response