feat: use internal ollama embeddings

harness/tht/cli/ollama_cmd.py

@@ -39,16 +39,16 @@ def _installed_models(base_url: str, timeout: float = 5.0) -> set[str]:
 
 def _start(start_cmd: list[str]) -> None:
     # Detached so the server outlives this short-lived CLI process.
-    subprocess.Popen(  # noqa: S603
+    subprocess.Popen(
         start_cmd, start_new_session=True,
         stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
     )
 
 
 def _warm(cfg) -> None:
-    from tht.vectorstore.embeddings import OllamaEmbeddings
+    from tht.vectorstore.embeddings import OllamaInternalEmbeddings
 
-    OllamaEmbeddings(cfg.embeddings).embed_query("ping")
+    OllamaInternalEmbeddings(cfg.embeddings).embed_query("ping")
 
 
 def _model_present(installed: set[str], model: str) -> bool:
@@ -104,7 +104,7 @@ def ensure_ollama(
                 if probe(base_url):
                     up = True
                     break
-            except Exception:  # noqa: BLE001 - transient during startup; keep polling
+            except Exception:  # noqa: BLE001,S112 - transient during startup; keep polling
                 continue
         if not up:
             return {"ok": False, "stage": "server",