deploy: route embedder through OpenAI-compat proxy instead of Ollama

The custom OpenAI-compatible endpoint (LiteLLM) serves the same
qwen3-embedding model and is reachable from the container in all
deployments; direct Ollama may not be. Vectors stay compatible because
the underlying model is the same.

Captured from a beast production hotfix.

backend/mem0_manager.py

@@ -102,10 +102,16 @@ class Mem0Manager:
                 },
             },
             "embedder": {
-                "provider": "ollama",
+                # Route embeddings through the OpenAI-compatible LiteLLM proxy
+                # rather than Ollama directly — the proxy is reachable from the
+                # container in all deployments, Ollama may not be. The model
+                # name is the same (qwen3-embedding:4b-q8_0); existing vectors
+                # generated via this path stay compatible.
+                "provider": "openai",
                 "config": {
                     "model": settings.embedding_model,
-                    "ollama_base_url": settings.ollama_base_url,
+                    "api_key": settings.openai_api_key,
+                    "openai_base_url": settings.openai_base_url,
                     "embedding_dims": settings.embedding_dims,
                 },
             },