fix: dashboard tolerates API cold starts (retry + friendly message instead of traceback)

Author: kohsheen1234

dashboard/app.py

@@ -1,4 +1,4 @@
-from dashboard.client import AgentWatchClient
+from dashboard.client import AgentWatchClient, APIUnavailable
 
 # Plain-language descriptions of each incident category, shown in the UI so a
 # first-time visitor understands what the labels mean.
@@ -182,12 +182,21 @@ def render() -> None:
     page = _sidebar(st)
     st.title("AgentWatch — AI Incident Observatory")
 
-    if page == "Overview":
-        _render_overview(st, api)
-    elif page == "Incident Explorer":
-        _render_explorer(st, api)
-    elif page == "Review Queue":
-        _render_review(st, api)
+    try:
+        if page == "Overview":
+            _render_overview(st, api)
+        elif page == "Incident Explorer":
+            _render_explorer(st, api)
+        elif page == "Review Queue":
+            _render_review(st, api)
+    except APIUnavailable:
+        st.warning(
+            "⏳ The API isn't responding yet. This demo runs on Render's free tier, "
+            "where services sleep after ~15 minutes of inactivity and take 30–60s to "
+            "wake up. Give it a moment and retry."
+        )
+        if st.button("Retry"):
+            st.rerun()
 
 
 if __name__ == "__main__":

dashboard/client.py

@@ -1,4 +1,9 @@
 import os
+import time
+
+
+class APIUnavailable(Exception):
+    """Raised when the API cannot be reached or returns a non-JSON response."""
 
 
 class AgentWatchClient:
@@ -10,28 +15,45 @@ def __init__(self, base_url: str | None = None, api_key: str | None = None, clie
         if client is None:
             import httpx
 
-            client = httpx.Client(base_url=self.base_url, timeout=30.0)
+            client = httpx.Client(base_url=self.base_url, timeout=30.0, follow_redirects=True)
         self._client = client
 
     def _headers(self) -> dict:
         return {"X-API-Key": self.api_key} if self.api_key else {}
 
+    def _get_json(self, path: str, params: dict | None = None, *, attempts: int = 3):
+        last: Exception | None = None
+        for i in range(attempts):
+            try:
+                resp = self._client.get(path, params=params or {})
+                resp.raise_for_status()
+                return resp.json()
+            except Exception as exc:  # network error, non-2xx, or non-JSON body
+                last = exc
+                if i < attempts - 1:
+                    time.sleep(1.5 * (i + 1))
+        raise APIUnavailable(str(last)) from last
+
     def incidents(self, **filters) -> dict:
         params = {k: v for k, v in filters.items() if v is not None}
-        return self._client.get("/incidents", params=params).json()
+        return self._get_json("/incidents", params)
 
     def incident(self, incident_id: int) -> dict:
-        return self._client.get(f"/incidents/{incident_id}").json()
+        return self._get_json(f"/incidents/{incident_id}")
 
     def stats(self) -> dict:
-        return self._client.get("/stats").json()
+        return self._get_json("/stats")
 
     def review(
         self, incident_id: int, *, reviewer: str, decision: str, notes: str | None = None
     ) -> dict:
-        resp = self._client.post(
-            f"/incidents/{incident_id}/review",
-            json={"reviewer": reviewer, "decision": decision, "notes": notes},
-            headers=self._headers(),
-        )
-        return resp.json()
+        try:
+            resp = self._client.post(
+                f"/incidents/{incident_id}/review",
+                json={"reviewer": reviewer, "decision": decision, "notes": notes},
+                headers=self._headers(),
+            )
+            resp.raise_for_status()
+            return resp.json()
+        except Exception as exc:
+            raise APIUnavailable(str(exc)) from exc

tests/test_dashboard_client.py

@@ -54,6 +54,29 @@ def _seed(monkeypatch, tmp_path):
         )
 
 
+def test_client_raises_apiunavailable_on_error(monkeypatch):
+    import pytest
+
+    import dashboard.client as c
+
+    monkeypatch.setattr(c.time, "sleep", lambda *a, **k: None)  # no waiting in tests
+
+    class BadResp:
+        def raise_for_status(self):
+            raise RuntimeError("502 Bad Gateway (cold start)")
+
+        def json(self):
+            return {}
+
+    class BadClient:
+        def get(self, *a, **k):
+            return BadResp()
+
+    api = c.AgentWatchClient(client=BadClient())
+    with pytest.raises(c.APIUnavailable):
+        api.stats()
+
+
 def test_client_reads_and_reviews(monkeypatch, tmp_path):
     _seed(monkeypatch, tmp_path)
     api = AgentWatchClient(client=TestClient(create_app()))