Upload folder using huggingface_hub

api/main.py

@@ -181,7 +181,7 @@ async def inspect(
         calibrated_score=result.calibrated_score,
         score_std=result.score_std,
         category=result.category,
-        model_version=MODEL_VERSION,
+        version=MODEL_VERSION,
         heatmap_b64=result.heatmap_b64,
         defect_crop_b64=result.defect_crop_b64,
         depth_map_b64=result.depth_map_b64,
@@ -501,7 +501,7 @@ async def health():
 
     return HealthResponse(
         status="ok",
-        model_version=MODEL_VERSION,
+        version=MODEL_VERSION,
         uptime_seconds=round(get_uptime(), 1),
         index_sizes=index_status,
         coreset_size=sum(

api/schemas.py

@@ -22,7 +22,7 @@ class InspectResponse(BaseModel):
     calibrated_score:   float = Field(..., ge=0.0, le=1.0)
     score_std:          float
     category:           str
-    model_version:      str
+    version:            str
 
     # Visuals (base64 PNG strings)
     heatmap_b64:        Optional[str] = None
@@ -116,7 +116,7 @@ class CorrectionResponse(BaseModel):
 # ── /health ──────────────────────────────────────────────────
 class HealthResponse(BaseModel):
     status:               str
-    model_version:        str
+    version:              str
     uptime_seconds:       float
     index_sizes:          dict
     coreset_size:         int

app.py

@@ -101,7 +101,7 @@ def run_inspector(image, category_hint, last_click_state):
     meta    = (f"Category: {category} | "
                f"Raw score: {result.get('anomaly_score', 0):.4f} | "
                f"Latency: {latency:.0f}ms | "
-               f"Model: {result.get('model_version', 'v1.0')}")
+               f"Model: {result.get('version', 'v1.0')}")
 
     # Poll LLM report
     report_id = result.get("report_id")

src/graph.py

@@ -24,21 +24,34 @@ class KnowledgeGraph:
     def load(self):
         path = os.path.join(self.data_dir, "knowledge_graph.json")
         if not os.path.exists(path):
-            raise FileNotFoundError(f"Knowledge graph not found: {path}")
+            print(f"Warning: Knowledge graph not found at {path}, using empty graph")
+            self.graph = nx.DiGraph()
+            return
 
-        with open(path) as f:
-            data = json.load(f)
-
-        # Handle both old and new NetworkX node-link formats
-        if "links" in data:
-            data["links"] = data.get("links", [])
         try:
-            self.graph = nx.node_link_graph(data)
-        except Exception:
-            # Try with edges key for older NetworkX
-            if "links" in data and "edges" not in data:
-                data["edges"] = data.pop("links")
-            self.graph = nx.node_link_graph(data)
+            with open(path) as f:
+                data = json.load(f)
+        except Exception as e:
+            print(f"Warning: Failed to load knowledge graph JSON: {e}")
+            self.graph = nx.DiGraph()
+            return
+
+        # Try standard NetworkX node_link_graph format (expects 'links' or 'edges')
+        try:
+            self.graph = nx.node_link_graph(data, directed=True)
+        except (KeyError, TypeError):
+            # If that fails, try converting 'edges' key to 'links'
+            if "edges" in data:
+                data["links"] = data.pop("edges")
+                try:
+                    self.graph = nx.node_link_graph(data, directed=True)
+                except Exception as e:
+                    print(f"Warning: Failed to load graph with edges→links conversion: {e}")
+                    self.graph = nx.DiGraph()
+            else:
+                # Last resort: create empty graph
+                print(f"Warning: Knowledge graph format not recognized, using empty graph")
+                self.graph = nx.DiGraph()
 
         print(f"Knowledge graph loaded: "
               f"{self.graph.number_of_nodes()} nodes, "