Cache validation logic fix

app.py

@@ -103,8 +103,12 @@ def fetch_project_data(proj, url, headers):
 
 
 @st.cache_data(ttl=300)
-def load_data():
-    """Load annotation data from Label Studio with disk cache."""
+def load_data(projects_hash):
+    """Load annotation data from Label Studio with disk cache.
+
+    Args:
+        projects_hash: Hash of project states to invalidate Streamlit cache when projects change
+    """
     try:
         url = st.secrets.get("LABEL_STUDIO_URL", os.getenv("LABEL_STUDIO_URL", "")).rstrip("/")
         key = st.secrets.get("LABEL_STUDIO_API_KEY", os.getenv("LABEL_STUDIO_API_KEY", ""))
@@ -191,6 +195,36 @@ def load_data():
     return df
 
 
+def get_projects_hash():
+    """Fetch projects and return a hash of their states for cache invalidation."""
+    import hashlib
+
+    try:
+        url = st.secrets.get("LABEL_STUDIO_URL", os.getenv("LABEL_STUDIO_URL", "")).rstrip("/")
+        key = st.secrets.get("LABEL_STUDIO_API_KEY", os.getenv("LABEL_STUDIO_API_KEY", ""))
+    except (KeyError, FileNotFoundError, AttributeError):
+        url = os.getenv("LABEL_STUDIO_URL", "").rstrip("/")
+        key = os.getenv("LABEL_STUDIO_API_KEY", "")
+
+    if not url or not key:
+        return "no-credentials"
+
+    headers = {"Authorization": f"Token {key}"}
+    resp = requests.get(f"{url}/api/projects", headers=headers, timeout=30)
+    resp.raise_for_status()
+    projects = resp.json().get("results", [])
+
+    # Create hash from project states (id, task_number, num_tasks_with_annotations)
+    state_string = ""
+    for proj in projects:
+        pid = proj["id"]
+        task_count = proj.get("task_number", 0)
+        submitted_count = proj.get("num_tasks_with_annotations", 0)
+        state_string += f"{pid}:{task_count}:{submitted_count};"
+
+    return hashlib.md5(state_string.encode()).hexdigest()
+
+
 def anonymize(name):
     """Convert '26 [Name Lastname]' to 'N.L. (26)'"""
     if name == "Others":
@@ -209,7 +243,8 @@ st.markdown("---")
 
 # Load data
 with st.spinner("Loading..."):
-    df = load_data()
+    projects_hash = get_projects_hash()
+    df = load_data(projects_hash)
 
 # Overview metrics
 total = df[df["is_goal_state"]]["words"].sum()