From b578637112d97f625734cbe48eebcd68cda8cf5d Mon Sep 17 00:00:00 2001
From: ChemaVX <jmivanez@gmail.com>
Date: Fri, 12 Jun 2026 08:12:20 +0000
Subject: [PATCH] feat(polymarket-bot): daily retention CronJob for
 metrics_daily

The bot snapshots metrics every cycle (~1,300 rows/day, 74K rows in 57
days) but only the last snapshot of each past UTC day is ever read back.
Prune past days to their daily close at 00:10 UTC; today's rows are kept
intact for /api/summary.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 polymarket-bot/cronjob-metrics-retention.yaml | 54 +++++++++++++++++++
 1 file changed, 54 insertions(+)
 create mode 100644 polymarket-bot/cronjob-metrics-retention.yaml

diff --git a/polymarket-bot/cronjob-metrics-retention.yaml b/polymarket-bot/cronjob-metrics-retention.yaml
new file mode 100644
index 0000000..804adec
--- /dev/null
+++ b/polymarket-bot/cronjob-metrics-retention.yaml
@@ -0,0 +1,54 @@
+apiVersion: batch/v1
+kind: CronJob
+metadata:
+  name: metrics-retention
+  namespace: polymarket-bot
+spec:
+  # Daily at 00:10 UTC: collapse every past UTC day in metrics_daily to its
+  # last snapshot (the "daily close"). The bot writes one snapshot per cycle
+  # (~1,300/day); only the close of past days is read back (DISTINCT ON in
+  # get_metrics_history / get_daily_pnl_closes), so intraday rows of past
+  # days are dead weight. Today's rows are never touched: the latest one
+  # feeds /api/summary and the close is still evolving.
+  schedule: "10 0 * * *"
+  concurrencyPolicy: Forbid
+  successfulJobsHistoryLimit: 3
+  failedJobsHistoryLimit: 3
+  jobTemplate:
+    spec:
+      backoffLimit: 2
+      template:
+        spec:
+          restartPolicy: Never
+          containers:
+            - name: prune
+              image: postgres:16-alpine
+              env:
+                - name: DATABASE_URL
+                  valueFrom:
+                    secretKeyRef:
+                      name: bot-secrets
+                      key: DATABASE_URL
+              command:
+                - /bin/sh
+                - -c
+                - |
+                  set -e
+                  psql "$DATABASE_URL" -v ON_ERROR_STOP=1 -c "
+                    DELETE FROM metrics_daily m
+                    USING (
+                        SELECT timestamp::date AS day, MAX(timestamp) AS keep_ts
+                        FROM metrics_daily
+                        WHERE timestamp::date < CURRENT_DATE
+                        GROUP BY 1
+                    ) k
+                    WHERE m.timestamp::date = k.day
+                      AND m.timestamp < k.keep_ts;
+                  "
+                  psql "$DATABASE_URL" -c "VACUUM ANALYZE metrics_daily;"
+              resources:
+                requests:
+                  cpu: 10m
+                  memory: 32Mi
+                limits:
+                  memory: 128Mi