feat: add cv benchmark workflow and admin visibility

tools/summarizer/README.md

@@ -34,7 +34,7 @@ python -m uvicorn app:app --host 127.0.0.1 --port 8001 --workers 1
 The Dockerfile installs Tesseract OCR so scanned PDFs and supported images can be processed inside the container.
 
 ## API
-- `GET /health` — health check and runtime capabilities
+- `GET /health` — health check and runtime capabilities, including Ollama version/model metadata when configured
 - `POST /summarize` — JSON body `{ "text": "...", "max_length": 150, "min_length": 30 }`
 - `POST /extract-text` — multipart file upload, returns extracted text and OCR metadata
 - `POST /cv/classify-block` — JSON body `{ "block": "..." }`, uses Ollama when `OLLAMA_MODEL` is configured

tools/summarizer/app.py

@@ -63,6 +63,12 @@ def _key(text: str, max_length: int, min_length: int, top_skills: int) -> str:
     return f"{h}:{max_length}:{min_length}:{top_skills}"
 
 
+def _ollama_json(path: str):
+    req = urllib_request.Request(f"{OLLAMA_BASE_URL}{path}", method="GET")
+    with urllib_request.urlopen(req, timeout=5) as response:
+        return json.loads(response.read().decode("utf-8"))
+
+
 def _ollama_status():
     configured = bool(OLLAMA_MODEL)
     if not configured:
@@ -71,27 +77,43 @@ def _ollama_status():
             "ollama_reachable": False,
             "ollama_model": None,
             "ollama_model_available": False,
+            "ollama_version": None,
+            "ollama_installed_models": [],
+            "ollama_loaded_models": [],
+            "ollama_loaded_count": 0,
         }
 
-    req = urllib_request.Request(f"{OLLAMA_BASE_URL}/api/tags", method="GET")
     try:
-        with urllib_request.urlopen(req, timeout=5) as response:
-            body = json.loads(response.read().decode("utf-8"))
+        tags_body = _ollama_json("/api/tags")
+        version_body = _ollama_json("/api/version")
+        try:
+            ps_body = _ollama_json("/api/ps")
+        except Exception:
+            ps_body = {"models": []}
     except Exception:
         return {
             "ollama_configured": True,
             "ollama_reachable": False,
             "ollama_model": OLLAMA_MODEL,
             "ollama_model_available": False,
+            "ollama_version": None,
+            "ollama_installed_models": [],
+            "ollama_loaded_models": [],
+            "ollama_loaded_count": 0,
         }
 
-    models = body.get("models") or []
-    names = {item.get("name") for item in models if isinstance(item, dict)}
+    models = tags_body.get("models") or []
+    names = sorted({item.get("name") for item in models if isinstance(item, dict) and item.get("name")})
+    loaded_models = sorted({item.get("name") for item in (ps_body.get("models") or []) if isinstance(item, dict) and item.get("name")})
     return {
         "ollama_configured": True,
         "ollama_reachable": True,
         "ollama_model": OLLAMA_MODEL,
         "ollama_model_available": OLLAMA_MODEL in names,
+        "ollama_version": version_body.get("version"),
+        "ollama_installed_models": names,
+        "ollama_loaded_models": loaded_models,
+        "ollama_loaded_count": len(loaded_models),
     }
 
 

tools/summarizer/tests/test_app.py

@@ -32,6 +32,8 @@ def test_health_reports_runtime_without_ollama(monkeypatch):
     assert payload["device"] == "cpu"
     assert payload["ollama_configured"] is False
     assert payload["ollama_model"] is None
+    assert payload["ollama_installed_models"] == []
+    assert payload["ollama_loaded_models"] == []
 
 
 def test_classify_block_returns_structured_json(monkeypatch):