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 +1,12 @@
+# Every dashboard query, run against Prometheus and through Grafana's query API after scripts/load.sh
+
+panel                                                      prom series grafana frames
+Cost per hour by endpoint (USD, illustrative pri [A]                 4              4
+Tokens per minute by endpoint and direction [A]                      8              8
+Cost per request by endpoint (USD) [A]                               4              4
+Model call latency p50 / p95 (s) [A]                                 1              1
+Model call latency p50 / p95 (s) [B]                                 1              1
+Failed model calls (share) [A]                                       1              1
+Tool calls per minute [A]                                            1              1
+Mean tool latency (s) [A]                                            1              1
+Calls with no cost recorded [A]                                      1              1