Add observability module: Spring AI built-in meters and spans, cost per endpoint from token usage, Prometheus and Grafana dashboard

Co-Authored-By: Claude Sonnet 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01JXVi2GMQ7bR5EmbUFdDj7N
2026-10-09 06:39:21 +00:00
parent 65f580c5bc
commit cc2a1b8bcb
45 changed files with 1731 additions and 0 deletions
@@ -0,0 +1,38 @@
+#!/usr/bin/env python3
+"""Runs every dashboard query twice: straight against Prometheus, and through Grafana's /api/ds/query with the
+provisioned datasource (what a panel really does). Prints one line per panel query and exits non-zero if any
+query errors or returns no series. Usage: verify-dashboard.py [output-file]"""
+import base64, json, pathlib, sys, time, urllib.parse, urllib.request
+
+root = pathlib.Path(__file__).resolve().parent.parent
+dash = json.loads((root / "dashboards" / "spring-ai-observability.json").read_text())
+auth = "Basic " + base64.b64encode(b"admin:admin").decode()
+
+def get(url, data=None, headers=None):
+    req = urllib.request.Request(url, data=data, headers=headers or {})
+    with urllib.request.urlopen(req, timeout=30) as r:
+        return json.loads(r.read())
+
+lines, bad = [], 0
+now = int(time.time())
+lines.append(f"{'panel':<58} {'prom series':>11} {'grafana frames':>14}")
+for p in dash["panels"]:
+    for t in p["targets"]:
+        q = urllib.parse.urlencode({"query": t["expr"]})
+        prom = get("http://127.0.0.1:9090/api/v1/query?" + q)
+        n_prom = len(prom["data"]["result"]) if prom["status"] == "success" else -1
+        body = json.dumps({"queries": [{"refId": t["refId"], "datasource": t["datasource"], "expr": t["expr"], "instant": False,
+                                        "range": True, "interval": "", "intervalMs": 5000, "maxDataPoints": 200}],
+                           "from": str((now - 600) * 1000), "to": str(now * 1000)}).encode()
+        gf = get("http://127.0.0.1:3000/api/ds/query", body, {"Content-Type": "application/json", "Authorization": auth})
+        res = gf["results"][t["refId"]]
+        n_gf = -1 if "error" in res else len(res.get("frames", []))
+        name = f"{p['title'][:48]} [{t['refId']}]"
+        lines.append(f"{name:<58} {n_prom:>11} {n_gf:>14}")
+        if n_prom < 1 or n_gf < 1:
+            bad += 1
+out = "\n".join(lines)
+print(out)
+if len(sys.argv) > 1:
+    pathlib.Path(sys.argv[1]).write_text("# Every dashboard query, run against Prometheus and through Grafana's query API after scripts/load.sh\n\n" + out + "\n")
+sys.exit(1 if bad else 0)