Enhance Ollama model management and performance tuning

- Updated the `provision_llm` function to utilize the `/api/tags` endpoint for verifying available models, improving accuracy in model management.
- Introduced a new `already_have_ollama_tag` function to ensure exact tag matching, preventing mismatches during model checks.
- Enhanced the `pull_stream` function to require a successful status from the API before proceeding, ensuring reliable model downloads.
- Added logic to handle unwritten blob files, improving the robustness of the model pulling process.
- Updated documentation and tests to reflect these changes, ensuring clarity and reliability in Ollama model operations.

tests/test_verify_stack.py

@@ -188,4 +188,59 @@ def test_stack_probe_requires_ollama_models():
 
     assert "/api/tags" in _REMOTE_STACK_PROBE
     assert "0 models" in _REMOTE_STACK_PROBE
-    assert 'payload.get("models")' in _REMOTE_STACK_PROBE
+    assert "WANT_OLLAMA_MODELS" in _REMOTE_STACK_PROBE
+    assert "retry" in _REMOTE_STACK_PROBE
+    assert ".gpu-rent-ollama-pulling" in _REMOTE_STACK_PROBE
+    assert "WARN 0 models" in _REMOTE_STACK_PROBE
+    assert "GPU не гасим" in _REMOTE_STACK_PROBE
+
+
+def test_verify_stack_zero_models_does_not_fail_up(monkeypatch):
+    from gpu_rent.ready import ServiceCheck, verify_stack_on_vm
+
+    class C:
+        enable_swarmui = True
+        llm_runtime = "ollama"
+
+    def fake_probe(_cfg, _host):
+        return [
+            ServiceCheck("swarmui", True, "ok", "vm", retry=False),
+            ServiceCheck(
+                "ollama",
+                True,
+                "WARN 0 models — Assistent empty (GPU не гасим)",
+                "vm",
+                retry=False,
+            ),
+        ]
+
+    monkeypatch.setattr("gpu_rent.ready._probe_vm_once", fake_probe)
+    out = verify_stack_on_vm(C(), "1.2.3.4", [].append, timeout=5.0, poll_every=0.01)
+    assert all(c.ok for c in out)
+
+
+def test_verify_stack_retries_while_pulling(monkeypatch):
+    from gpu_rent.ready import ServiceCheck, verify_stack_on_vm
+
+    class C:
+        enable_swarmui = False
+        llm_runtime = "ollama"
+
+    n = {"i": 0}
+
+    def fake_probe(_cfg, _host):
+        n["i"] += 1
+        if n["i"] == 1:
+            return [
+                ServiceCheck(
+                    "ollama", False, "0 models — pull идёт (12s)", "vm", retry=True
+                )
+            ]
+        return [ServiceCheck("ollama", True, "1 models (qwen)", "vm", retry=False)]
+
+    monkeypatch.setattr("gpu_rent.ready._probe_vm_once", fake_probe)
+    out = verify_stack_on_vm(
+        C(), "1.2.3.4", [].append, timeout=5.0, poll_every=0.01
+    )
+    assert out[0].ok
+    assert n["i"] == 2