Fix dashboard rendering: barchart panel type + metadata off the table prefix

Two issues found loading the deployed dashboard:

1. Panels used type "bar-chart" (hyphenated); Grafana's core panel is "barchart"
   — hence "plugin bar-chart required". Fixed both panels.

2. ALL panels showed "no data" because Athena failed with HIVE_BAD_DATA:
   the classifier wrote summary.json/details.json INTO analytics/dt=*/ — the
   same prefix the Glue table scans — so Athena tried to read the JSON as
   Parquet and every query failed. Move the metadata to a separate meta/dt=*/
   prefix: classifier writes there (grant_read_write meta/*), the Slack Lambdas
   read there (read_meta_json, grant_read meta/*), and analytics/ holds only
   Parquet. Verified: the category GROUP BY query now succeeds against Athena.

cdk/stacks/pipeline_stack.py

@@ -274,6 +274,9 @@ class PipelineStack(Stack):
         # Parquet to S3 — it never touches the catalog (Phase 3).
         self.exports_bucket.grant_read(self.classifier_fn, "raw/*")
         self.exports_bucket.grant_read_write(self.classifier_fn, "analytics/*")
+        # summary.json/details.json live under meta/ (kept out of the Athena
+        # table's analytics/ prefix so queries don't read JSON as Parquet).
+        self.exports_bucket.grant_read_write(self.classifier_fn, "meta/*")
         secretsmanager.Secret.from_secret_name_v2(
             self, "AnthropicKey", ANTHROPIC_SECRET
         ).grant_read(self.classifier_fn)
@@ -335,7 +338,8 @@ class PipelineStack(Stack):
                     ),
                 ),
             )
-            self.exports_bucket.grant_read(fn, "analytics/*")
+            # Slack Lambdas read only the daily JSON under meta/ (not the Parquet).
+            self.exports_bucket.grant_read(fn, "meta/*")
             slack_secret.grant_read(fn)
             dashboard_param.grant_read(fn)
             return fn

grafana/dashboards/apm-work-orders.json

@@ -179,7 +179,7 @@
   "panels": [
     {
       "id": 1,
-      "type": "bar-chart",
+      "type": "barchart",
       "title": "Category Distribution",
       "description": "Count of work orders by classification category for the selected snapshot date. Sorted descending by volume. Filter using the template variables above.",
       "gridPos": { "x": 0, "y": 0, "w": 14, "h": 9 },
@@ -366,7 +366,7 @@
     },
     {
       "id": 4,
-      "type": "bar-chart",
+      "type": "barchart",
       "title": "Escalations by Site",
       "description": "Total escalations per site for the selected snapshot date. Includes all escalation categories.",
       "gridPos": { "x": 8, "y": 9, "w": 16, "h": 8 },

lambdas/classifier/handler.py

@@ -30,7 +30,8 @@ import pandas as pd
 
 import classify as clf
 
-ANALYTICS_PREFIX = "analytics"
+ANALYTICS_PREFIX = "analytics"  # Parquet snapshots — the Glue/Athena table reads this
+META_PREFIX = "meta"  # summary.json/details.json — kept OUT of the table's prefix
 
 # Map snapshot field -> substring matched (case-insensitively) against the export
 # header, tolerating minor header drift in the 13-column APM export.
@@ -237,17 +238,19 @@ def handler(event, context):
         mode="overwrite_partitions",
     )
 
+    # summary.json/details.json go under meta/ — NOT analytics/. Athena reads
+    # every object in the table's prefix as Parquet, so JSON there breaks queries.
     summary = _build_summary(df, dt, key, blank)
     _s3.put_object(
         Bucket=bucket,
-        Key=f"{ANALYTICS_PREFIX}/dt={dt}/summary.json",
+        Key=f"{META_PREFIX}/dt={dt}/summary.json",
         Body=json.dumps(summary, indent=2).encode("utf-8"),
         ContentType="application/json",
     )
     # details.json — per-WO index the slack-post Lambda reads for drill-down modals.
     _s3.put_object(
         Bucket=bucket,
-        Key=f"{ANALYTICS_PREFIX}/dt={dt}/details.json",
+        Key=f"{META_PREFIX}/dt={dt}/details.json",
         Body=json.dumps(_build_details(df)).encode("utf-8"),
         ContentType="application/json",
     )

lambdas/slack_post/handler.py

@@ -25,11 +25,11 @@ def handler(event, context):
     """Post the daily summary and (conditionally) the 3rd-escalation alert."""
     dt = (event or {}).get("dt") or datetime.now(timezone.utc).strftime("%Y-%m-%d")
 
-    today = slackio.read_analytics_json(dt, "summary.json")
+    today = slackio.read_meta_json(dt, "summary.json")
     if today is None:
         print(f"No summary.json for dt={dt}; nothing to post.")
         return {"posted": False, "reason": "no summary", "dt": dt}
-    yesterday = slackio.read_analytics_json(_yesterday(dt), "summary.json")
+    yesterday = slackio.read_meta_json(_yesterday(dt), "summary.json")
 
     client = slackio.web_client()
     channel = slackio.channel_id()
@@ -44,7 +44,7 @@ def handler(event, context):
     # Standalone batched alert — only when there are 3rd escalations. Pull the
     # rows from details.json so the alert can name the WOs.
     if today.get("third_escalation_count", 0) > 0:
-        details = slackio.read_analytics_json(dt, "details.json") or []
+        details = slackio.read_meta_json(dt, "details.json") or []
         thirds = [d for d in details if d.get("category") == "3rd Escalation"]
         alert_blocks = blockkit.build_escalation_alert(thirds)
         if alert_blocks:

lambdas/slack_post/interactions.py

@@ -60,7 +60,7 @@ def handler(event, context):
 
     # The daily post is same-day; default the drill to today's snapshot.
     dt = datetime.now(timezone.utc).strftime("%Y-%m-%d")
-    details = slackio.read_analytics_json(dt, "details.json") or []
+    details = slackio.read_meta_json(dt, "details.json") or []
     wos = [d for d in details if d.get(field) == value]
 
     view = blockkit.build_wo_modal(

lambdas/slack_post/slackio.py

@@ -60,9 +60,10 @@ def verify_signature(body: str, timestamp: str, signature: str) -> bool:
     return verifier.is_valid(body=body, timestamp=timestamp, signature=signature)
 
 
-def read_analytics_json(dt: str, name: str):
-    """Read ``analytics/dt=<dt>/<name>`` as JSON, or None if absent."""
-    key = f"analytics/dt={dt}/{name}"
+def read_meta_json(dt: str, name: str):
+    """Read ``meta/dt=<dt>/<name>`` as JSON, or None if absent. (summary.json /
+    details.json live under meta/, kept out of the Athena table's analytics/ prefix.)"""
+    key = f"meta/dt={dt}/{name}"
     try:
         obj = _s3.get_object(Bucket=_BUCKET, Key=key)
     except _s3.exceptions.NoSuchKey: