diff --git a/apps/server/otel/grafana/dashboards/airi-server-overview-cloud.json b/apps/server/otel/grafana/dashboards/airi-server-overview-cloud.json
index f0226c83a..fd6df97b1 100644
--- a/apps/server/otel/grafana/dashboards/airi-server-overview-cloud.json
+++ b/apps/server/otel/grafana/dashboards/airi-server-overview-cloud.json
@@ -48,7 +48,7 @@
       "type": "stat",
       "targets": [
         {
-          "expr": "sum(rate(http_server_request_duration_milliseconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
+          "expr": "sum(rate(http_server_request_duration_seconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
           "legendFormat": "req/s"
         }
       ]
@@ -80,7 +80,7 @@
       "type": "stat",
       "targets": [
         {
-          "expr": "sum(rate(http_server_request_duration_milliseconds_count{service_name=~\"$service\", deployment_environment=~\"$env\", http_status_code=~\"5..\"}[$__rate_interval])) / sum(rate(http_server_request_duration_milliseconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
+          "expr": "sum(rate(http_server_request_duration_seconds_count{service_name=~\"$service\", deployment_environment=~\"$env\", http_status_code=~\"5..\"}[$__rate_interval])) / sum(rate(http_server_request_duration_seconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
           "legendFormat": "error rate"
         }
       ]
@@ -94,11 +94,11 @@
             "mode": "absolute",
             "steps": [
               { "color": "green", "value": null },
-              { "color": "yellow", "value": 200 },
-              { "color": "red", "value": 1000 }
+              { "color": "yellow", "value": 0.2 },
+              { "color": "red", "value": 1.0 }
             ]
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 8, "x": 16, "y": 1 },
@@ -108,11 +108,11 @@
         "colorMode": "value",
         "graphMode": "area"
       },
-      "title": "P95 Latency",
+      "title": "P95 Latency (excl. LLM)",
       "type": "stat",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum(rate(http_server_request_duration_milliseconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval])) by (le))",
+          "expr": "histogram_quantile(0.95, sum(rate(http_server_request_duration_seconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\", http_route!~\"/api/v1/openai/.*\"}[$__rate_interval])) by (le))",
           "legendFormat": "p95"
         }
       ]
@@ -136,7 +136,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (http_route) (rate(http_server_request_duration_milliseconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
+          "expr": "sum by (http_route) (rate(http_server_request_duration_seconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
           "legendFormat": "{{http_route}}"
         }
       ]
@@ -150,7 +150,7 @@
             "lineInterpolation": "smooth",
             "fillOpacity": 20
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 12, "x": 12, "y": 9 },
@@ -159,7 +159,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum by (le, http_route) (rate(http_server_request_duration_milliseconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval])))",
+          "expr": "histogram_quantile(0.95, sum by (le, http_route) (rate(http_server_request_duration_seconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval])))",
           "legendFormat": "{{http_route}}"
         }
       ]
@@ -182,7 +182,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (http_status_code) (rate(http_server_request_duration_milliseconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
+          "expr": "sum by (http_status_code) (rate(http_server_request_duration_seconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
           "legendFormat": "{{http_status_code}}"
         }
       ]
@@ -318,7 +318,7 @@
             "lineInterpolation": "smooth",
             "fillOpacity": 20
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 12, "x": 0, "y": 43 },
@@ -327,7 +327,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum by (le, db_operation) (rate(db_client_operation_duration_milliseconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval])))",
+          "expr": "histogram_quantile(0.95, sum by (le, db_operation) (rate(db_client_operation_duration_seconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval])))",
           "legendFormat": "{{db_operation}}"
         }
       ]
@@ -350,7 +350,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (db_operation) (rate(db_client_operation_duration_milliseconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
+          "expr": "sum by (db_operation) (rate(db_client_operation_duration_seconds_count{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
           "legendFormat": "{{db_operation}}"
         }
       ]
@@ -402,7 +402,7 @@
             "lineInterpolation": "smooth",
             "fillOpacity": 20
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 12, "x": 0, "y": 60 },
@@ -411,7 +411,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum by (le, db_operation) (rate(db_client_operation_duration_milliseconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\", db_system=\"redis\"}[$__rate_interval])))",
+          "expr": "histogram_quantile(0.95, sum by (le, db_operation) (rate(db_client_operation_duration_seconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\", db_system=\"redis\"}[$__rate_interval])))",
           "legendFormat": "{{db_operation}}"
         }
       ]
@@ -434,7 +434,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (db_operation) (rate(db_client_operation_duration_milliseconds_count{service_name=~\"$service\", deployment_environment=~\"$env\", db_system=\"redis\"}[$__rate_interval]))",
+          "expr": "sum by (db_operation) (rate(db_client_operation_duration_seconds_count{service_name=~\"$service\", deployment_environment=~\"$env\", db_system=\"redis\"}[$__rate_interval]))",
           "legendFormat": "{{db_operation}}"
         }
       ]
@@ -473,7 +473,7 @@
       "type": "stat",
       "targets": [
         {
-          "expr": "sum(rate(llm_request_count_total{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
+          "expr": "sum(rate(gen_ai_client_operation_count_total{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
           "legendFormat": "req/s"
         }
       ]
@@ -487,11 +487,11 @@
             "mode": "absolute",
             "steps": [
               { "color": "green", "value": null },
-              { "color": "yellow", "value": 500 },
-              { "color": "red", "value": 2000 }
+              { "color": "yellow", "value": 0.5 },
+              { "color": "red", "value": 2.0 }
             ]
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 8, "x": 8, "y": 69 },
@@ -505,7 +505,7 @@
       "type": "stat",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum(rate(llm_request_duration_milliseconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval])) by (le))",
+          "expr": "histogram_quantile(0.95, sum(rate(gen_ai_client_operation_duration_seconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval])) by (le))",
           "legendFormat": "p95"
         }
       ]
@@ -537,7 +537,7 @@
       "type": "stat",
       "targets": [
         {
-          "expr": "sum(rate(flux_consumed_total{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
+          "expr": "sum(rate(airi_billing_flux_consumed_total{service_name=~\"$service\", deployment_environment=~\"$env\"}[$__rate_interval]))",
           "legendFormat": "flux/s"
         }
       ]
@@ -561,7 +561,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (gen_ai_request_model) (rate(llm_request_count_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval]))",
+          "expr": "sum by (gen_ai_request_model) (rate(gen_ai_client_operation_count_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval]))",
           "legendFormat": "{{gen_ai_request_model}}"
         }
       ]
@@ -584,11 +584,11 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (gen_ai_operation_name) (rate(llm_request_count_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_operation_name!=\"\"}[$__rate_interval]))",
+          "expr": "sum by (gen_ai_operation_name) (rate(gen_ai_client_operation_count_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_operation_name!=\"\"}[$__rate_interval]))",
           "legendFormat": "{{gen_ai_operation_name}}"
         },
         {
-          "expr": "sum by (airi_gen_ai_operation_kind) (rate(llm_request_count_total{service_name=~\"$service\", deployment_environment=~\"$env\", airi_gen_ai_operation_kind!=\"\"}[$__rate_interval]))",
+          "expr": "sum by (airi_gen_ai_operation_kind) (rate(gen_ai_client_operation_count_total{service_name=~\"$service\", deployment_environment=~\"$env\", airi_gen_ai_operation_kind!=\"\"}[$__rate_interval]))",
           "legendFormat": "{{airi_gen_ai_operation_kind}}"
         }
       ]
@@ -602,7 +602,7 @@
             "lineInterpolation": "smooth",
             "fillOpacity": 20
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 12, "x": 0, "y": 85 },
@@ -611,7 +611,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum by (le, gen_ai_request_model) (rate(llm_request_duration_milliseconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval])))",
+          "expr": "histogram_quantile(0.95, sum by (le, gen_ai_request_model) (rate(gen_ai_client_operation_duration_seconds_bucket{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval])))",
           "legendFormat": "{{gen_ai_request_model}}"
         }
       ]
@@ -634,7 +634,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (gen_ai_request_model) (rate(flux_consumed_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval]))",
+          "expr": "sum by (gen_ai_request_model) (rate(airi_billing_flux_consumed_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval]))",
           "legendFormat": "{{gen_ai_request_model}}"
         }
       ]
@@ -657,7 +657,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (gen_ai_request_model) (rate(llm_tokens_prompt_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval]))",
+          "expr": "sum by (gen_ai_request_model) (rate(gen_ai_client_token_usage_input_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval]))",
           "legendFormat": "{{gen_ai_request_model}}"
         }
       ]
@@ -680,7 +680,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (gen_ai_request_model) (rate(llm_tokens_completion_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval]))",
+          "expr": "sum by (gen_ai_request_model) (rate(gen_ai_client_token_usage_output_total{service_name=~\"$service\", deployment_environment=~\"$env\", gen_ai_request_model!=\"\"}[$__rate_interval]))",
           "legendFormat": "{{gen_ai_request_model}}"
         }
       ]
@@ -722,24 +722,24 @@
       {
         "current": { "selected": false, "text": "production", "value": "production" },
         "datasource": { "type": "prometheus", "uid": "grafanacloud-projairi-prom" },
-        "definition": "label_values(http_server_request_duration_milliseconds_count, deployment_environment)",
+        "definition": "label_values(http_server_request_duration_seconds_count, deployment_environment)",
         "hide": 0,
         "includeAll": true,
         "multi": false,
         "name": "env",
-        "query": "label_values(http_server_request_duration_milliseconds_count, deployment_environment)",
+        "query": "label_values(http_server_request_duration_seconds_count, deployment_environment)",
         "refresh": 2,
         "type": "query"
       },
       {
         "current": { "selected": false, "text": "server", "value": "server" },
         "datasource": { "type": "prometheus", "uid": "grafanacloud-projairi-prom" },
-        "definition": "label_values(http_server_request_duration_milliseconds_count{deployment_environment=~\"$env\"}, service_name)",
+        "definition": "label_values(http_server_request_duration_seconds_count{deployment_environment=~\"$env\"}, service_name)",
         "hide": 0,
         "includeAll": true,
         "multi": true,
         "name": "service",
-        "query": "label_values(http_server_request_duration_milliseconds_count{deployment_environment=~\"$env\"}, service_name)",
+        "query": "label_values(http_server_request_duration_seconds_count{deployment_environment=~\"$env\"}, service_name)",
         "refresh": 2,
         "type": "query"
       }
diff --git a/apps/server/otel/grafana/dashboards/airi-server-overview.json b/apps/server/otel/grafana/dashboards/airi-server-overview.json
index 6f9b4f5e9..8e894a3a7 100644
--- a/apps/server/otel/grafana/dashboards/airi-server-overview.json
+++ b/apps/server/otel/grafana/dashboards/airi-server-overview.json
@@ -48,7 +48,7 @@
       "type": "stat",
       "targets": [
         {
-          "expr": "sum(rate(airi_http_server_request_duration_milliseconds_count{service_name=~\"$service\"}[$__rate_interval]))",
+          "expr": "sum(rate(airi_http_server_request_duration_seconds_count{service_name=~\"$service\"}[$__rate_interval]))",
           "legendFormat": "req/s"
         }
       ]
@@ -80,7 +80,7 @@
       "type": "stat",
       "targets": [
         {
-          "expr": "sum(rate(airi_http_server_request_duration_milliseconds_count{service_name=~\"$service\", http_status_code=~\"5..\"}[$__rate_interval])) / sum(rate(airi_http_server_request_duration_milliseconds_count{service_name=~\"$service\"}[$__rate_interval]))",
+          "expr": "sum(rate(airi_http_server_request_duration_seconds_count{service_name=~\"$service\", http_status_code=~\"5..\"}[$__rate_interval])) / sum(rate(airi_http_server_request_duration_seconds_count{service_name=~\"$service\"}[$__rate_interval]))",
           "legendFormat": "error rate"
         }
       ]
@@ -94,11 +94,11 @@
             "mode": "absolute",
             "steps": [
               { "color": "green", "value": null },
-              { "color": "yellow", "value": 200 },
-              { "color": "red", "value": 1000 }
+              { "color": "yellow", "value": 0.2 },
+              { "color": "red", "value": 1.0 }
             ]
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 8, "x": 16, "y": 1 },
@@ -108,11 +108,11 @@
         "colorMode": "value",
         "graphMode": "area"
       },
-      "title": "P95 Latency",
+      "title": "P95 Latency (excl. LLM)",
       "type": "stat",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum(rate(airi_http_server_request_duration_milliseconds_bucket{service_name=~\"$service\"}[$__rate_interval])) by (le))",
+          "expr": "histogram_quantile(0.95, sum(rate(airi_http_server_request_duration_seconds_bucket{service_name=~\"$service\", http_route!~\"/api/v1/openai/.*\"}[$__rate_interval])) by (le))",
           "legendFormat": "p95"
         }
       ]
@@ -136,7 +136,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (http_route) (rate(airi_http_server_request_duration_milliseconds_count{service_name=~\"$service\"}[$__rate_interval]))",
+          "expr": "sum by (http_route) (rate(airi_http_server_request_duration_seconds_count{service_name=~\"$service\"}[$__rate_interval]))",
           "legendFormat": "{{http_route}}"
         }
       ]
@@ -150,7 +150,7 @@
             "lineInterpolation": "smooth",
             "fillOpacity": 20
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 12, "x": 12, "y": 9 },
@@ -159,7 +159,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum by (le, http_route) (rate(airi_http_server_request_duration_milliseconds_bucket{service_name=~\"$service\"}[$__rate_interval])))",
+          "expr": "histogram_quantile(0.95, sum by (le, http_route) (rate(airi_http_server_request_duration_seconds_bucket{service_name=~\"$service\"}[$__rate_interval])))",
           "legendFormat": "{{http_route}}"
         }
       ]
@@ -182,7 +182,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (http_status_code) (rate(airi_http_server_request_duration_milliseconds_count{service_name=~\"$service\"}[$__rate_interval]))",
+          "expr": "sum by (http_status_code) (rate(airi_http_server_request_duration_seconds_count{service_name=~\"$service\"}[$__rate_interval]))",
           "legendFormat": "{{http_status_code}}"
         }
       ]
@@ -318,7 +318,7 @@
             "lineInterpolation": "smooth",
             "fillOpacity": 20
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 12, "x": 0, "y": 43 },
@@ -327,7 +327,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum by (le, db_operation) (rate(airi_db_client_operation_duration_milliseconds_bucket{service_name=~\"$service\"}[$__rate_interval])))",
+          "expr": "histogram_quantile(0.95, sum by (le, db_operation) (rate(airi_db_client_operation_duration_seconds_bucket{service_name=~\"$service\"}[$__rate_interval])))",
           "legendFormat": "{{db_operation}}"
         }
       ]
@@ -350,7 +350,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (db_operation) (rate(airi_db_client_operation_duration_milliseconds_count{service_name=~\"$service\"}[$__rate_interval]))",
+          "expr": "sum by (db_operation) (rate(airi_db_client_operation_duration_seconds_count{service_name=~\"$service\"}[$__rate_interval]))",
           "legendFormat": "{{db_operation}}"
         }
       ]
@@ -402,7 +402,7 @@
             "lineInterpolation": "smooth",
             "fillOpacity": 20
           },
-          "unit": "ms"
+          "unit": "s"
         }
       },
       "gridPos": { "h": 8, "w": 12, "x": 0, "y": 60 },
@@ -411,7 +411,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "histogram_quantile(0.95, sum by (le, db_operation) (rate(airi_db_client_operation_duration_milliseconds_bucket{service_name=~\"$service\", db_system=\"redis\"}[$__rate_interval])))",
+          "expr": "histogram_quantile(0.95, sum by (le, db_operation) (rate(airi_db_client_operation_duration_seconds_bucket{service_name=~\"$service\", db_system=\"redis\"}[$__rate_interval])))",
           "legendFormat": "{{db_operation}}"
         }
       ]
@@ -434,7 +434,7 @@
       "type": "timeseries",
       "targets": [
         {
-          "expr": "sum by (db_operation) (rate(airi_db_client_operation_duration_milliseconds_count{service_name=~\"$service\", db_system=\"redis\"}[$__rate_interval]))",
+          "expr": "sum by (db_operation) (rate(airi_db_client_operation_duration_seconds_count{service_name=~\"$service\", db_system=\"redis\"}[$__rate_interval]))",
           "legendFormat": "{{db_operation}}"
         }
       ]
@@ -486,12 +486,12 @@
       {
         "current": { "selected": false, "text": "airi-server", "value": "airi-server" },
         "datasource": { "type": "prometheus", "uid": "${datasource}" },
-        "definition": "label_values(airi_http_server_request_duration_milliseconds_count, service_name)",
+        "definition": "label_values(airi_http_server_request_duration_seconds_count, service_name)",
         "hide": 0,
         "includeAll": true,
         "multi": true,
         "name": "service",
-        "query": "label_values(airi_http_server_request_duration_milliseconds_count, service_name)",
+        "query": "label_values(airi_http_server_request_duration_seconds_count, service_name)",
         "refresh": 2,
         "type": "query"
       }
diff --git a/apps/server/src/app.ts b/apps/server/src/app.ts
index d78a611fb..2121f849f 100644
--- a/apps/server/src/app.ts
+++ b/apps/server/src/app.ts
@@ -168,7 +168,7 @@ async function buildApp(deps: AppDeps) {
     /**
      * V1 routes for official provider.
      */
-    .route('/api/v1/openai', createV1CompletionsRoutes(deps.fluxService, deps.billingService, deps.configKV, deps.billingMq, deps.otel?.llm))
+    .route('/api/v1/openai', createV1CompletionsRoutes(deps.fluxService, deps.billingService, deps.configKV, deps.billingMq, deps.otel?.genAi))
 
     /**
      * Flux routes.
diff --git a/apps/server/src/libs/auth.ts b/apps/server/src/libs/auth.ts
index cd9ba3d8a..9b88c5d88 100644
--- a/apps/server/src/libs/auth.ts
+++ b/apps/server/src/libs/auth.ts
@@ -4,6 +4,7 @@ import type { AuthMetrics } from './otel'
 
 import { betterAuth } from 'better-auth'
 import { drizzleAdapter } from 'better-auth/adapters/drizzle'
+import { createAuthMiddleware } from 'better-auth/api'
 import { bearer } from 'better-auth/plugins'
 
 import { getAuthTrustedOrigins } from '../utils/origin'
@@ -53,6 +54,32 @@ export function createAuth(db: Database, env: Env, metrics?: AuthMetrics | null)
       },
     },
 
+    hooks: {
+      before: createAuthMiddleware(async (ctx) => {
+        const isAuthAttempt = ctx.path.includes('/sign-in') || ctx.path.includes('/sign-up')
+        if (isAuthAttempt) {
+          metrics?.attempts.add(1, { 'auth.method': ctx.path.split('/').pop() ?? 'unknown' })
+        }
+      }),
+      after: createAuthMiddleware(async (ctx) => {
+        // Track auth failures via otel
+        const isAuthAttempt = ctx.path.includes('/sign-in') || ctx.path.includes('/sign-up')
+        if (isAuthAttempt && ctx.context.returned && typeof ctx.context.returned === 'object' && 'error' in ctx.context.returned) {
+          metrics?.failures.add(1, { 'auth.method': ctx.path.split('/').pop() ?? 'unknown' })
+        }
+
+        // On OAuth callback errors, redirect back to the referer instead of returning API JSON
+        if (ctx.path.startsWith('/callback') && ctx.context.returned && typeof ctx.context.returned === 'object' && 'error' in ctx.context.returned) {
+          const referer = ctx.getHeader('referer')
+          if (referer) {
+            const url = new URL(referer)
+            url.searchParams.set('error', 'auth_failed')
+            throw ctx.redirect(url.toString())
+          }
+        }
+      }),
+    },
+
     databaseHooks: {
       user: {
         create: {
diff --git a/apps/server/src/libs/otel.ts b/apps/server/src/libs/otel.ts
index 6dd1f29db..8aca9efe9 100644
--- a/apps/server/src/libs/otel.ts
+++ b/apps/server/src/libs/otel.ts
@@ -23,6 +23,34 @@ import { NodeSDK } from '@opentelemetry/sdk-node'
 import { BatchSpanProcessor, ParentBasedSampler, TraceIdRatioBasedSampler } from '@opentelemetry/sdk-trace-node'
 import { ATTR_SERVICE_NAME, ATTR_SERVICE_VERSION } from '@opentelemetry/semantic-conventions'
 
+import {
+  METRIC_AUTH_ATTEMPTS,
+  METRIC_AUTH_FAILURES,
+  METRIC_CHARACTER_CREATED,
+  METRIC_CHARACTER_DELETED,
+  METRIC_CHARACTER_ENGAGEMENT,
+  METRIC_CHAT_MESSAGES,
+  METRIC_FLUX_CONSUMED,
+  METRIC_FLUX_INSUFFICIENT_BALANCE,
+  METRIC_GEN_AI_CLIENT_OPERATION_COUNT,
+  METRIC_GEN_AI_CLIENT_OPERATION_DURATION,
+  METRIC_GEN_AI_CLIENT_TOKEN_USAGE_INPUT,
+  METRIC_GEN_AI_CLIENT_TOKEN_USAGE_OUTPUT,
+  METRIC_HTTP_SERVER_ACTIVE_REQUESTS,
+  METRIC_HTTP_SERVER_REQUEST_DURATION,
+  METRIC_STRIPE_CHECKOUT_COMPLETED,
+  METRIC_STRIPE_CHECKOUT_CREATED,
+  METRIC_STRIPE_EVENTS,
+  METRIC_STRIPE_PAYMENT_FAILED,
+  METRIC_STRIPE_SUBSCRIPTION_EVENT,
+  METRIC_USER_ACTIVE_SESSIONS,
+  METRIC_USER_LOGIN,
+  METRIC_USER_REGISTERED,
+  METRIC_WS_CONNECTIONS_ACTIVE,
+  METRIC_WS_MESSAGES_RECEIVED,
+  METRIC_WS_MESSAGES_SENT,
+} from '../utils/observability'
+
 const logger = useLogger('otel')
 
 export interface HttpMetrics {
@@ -57,18 +85,18 @@ export interface RevenueMetrics {
   fluxInsufficientBalance: Counter
 }
 
-export interface LlmMetrics {
-  requestDuration: Histogram
-  requestCount: Counter
-  tokensPrompt: Counter
-  tokensCompletion: Counter
+export interface GenAiMetrics {
+  operationDuration: Histogram
+  operationCount: Counter
+  tokenUsageInput: Counter
+  tokenUsageOutput: Counter
   fluxConsumed: Counter
 }
 
-export interface DbMetrics {
-  queryDuration: Histogram
-  redisCommandDuration: Histogram
-}
+// NOTICE: Database metrics (db.client.operation.duration, redis.client.command.duration) were
+// intentionally removed. PgInstrumentation and IORedisInstrumentation already generate spans
+// with timing for every query/command. To surface these as metrics in Grafana, configure the
+// OTel Collector's spanmetrics connector to derive metrics from those spans.
 
 export interface OtelInstance {
   sdk: NodeSDK
@@ -76,8 +104,7 @@ export interface OtelInstance {
   auth: AuthMetrics
   engagement: EngagementMetrics
   revenue: RevenueMetrics
-  llm: LlmMetrics
-  db: DbMetrics
+  genAi: GenAiMetrics
   shutdown: () => Promise<void>
 }
 
@@ -168,115 +195,103 @@ export function initOtel(env: Env): OtelInstance | undefined {
 
   const meter = metrics.getMeter(serviceName)
 
-  // HTTP metrics
+  // HTTP metrics (semconv: unit MUST be seconds)
   const http: HttpMetrics = {
-    requestDuration: meter.createHistogram('http.server.request.duration', {
-      description: 'HTTP server request duration in milliseconds',
-      unit: 'ms',
+    requestDuration: meter.createHistogram(METRIC_HTTP_SERVER_REQUEST_DURATION, {
+      description: 'HTTP server request duration',
+      unit: 's',
     }),
-    activeRequests: meter.createUpDownCounter('http.server.active_requests', {
+    activeRequests: meter.createUpDownCounter(METRIC_HTTP_SERVER_ACTIVE_REQUESTS, {
       description: 'Number of active HTTP requests',
     }),
   }
 
   // Auth & User metrics
   const auth: AuthMetrics = {
-    attempts: meter.createCounter('auth.attempts', {
+    attempts: meter.createCounter(METRIC_AUTH_ATTEMPTS, {
       description: 'Number of authentication attempts',
     }),
-    failures: meter.createCounter('auth.failures', {
+    failures: meter.createCounter(METRIC_AUTH_FAILURES, {
       description: 'Number of failed authentication attempts',
     }),
-    userRegistered: meter.createCounter('user.registered', {
+    userRegistered: meter.createCounter(METRIC_USER_REGISTERED, {
       description: 'Number of new user registrations',
     }),
-    userLogin: meter.createCounter('user.login', {
+    userLogin: meter.createCounter(METRIC_USER_LOGIN, {
       description: 'Number of user logins',
     }),
-    activeSessions: meter.createUpDownCounter('user.active_sessions', {
+    activeSessions: meter.createUpDownCounter(METRIC_USER_ACTIVE_SESSIONS, {
       description: 'Number of active user sessions',
     }),
   }
 
   // Engagement metrics
   const engagement: EngagementMetrics = {
-    chatMessages: meter.createCounter('chat.messages', {
+    chatMessages: meter.createCounter(METRIC_CHAT_MESSAGES, {
       description: 'Number of chat messages written or pulled',
     }),
-    characterCreated: meter.createCounter('character.created', {
+    characterCreated: meter.createCounter(METRIC_CHARACTER_CREATED, {
       description: 'Number of characters created',
     }),
-    characterDeleted: meter.createCounter('character.deleted', {
+    characterDeleted: meter.createCounter(METRIC_CHARACTER_DELETED, {
       description: 'Number of characters deleted',
     }),
-    characterEngagement: meter.createCounter('character.engagement', {
+    characterEngagement: meter.createCounter(METRIC_CHARACTER_ENGAGEMENT, {
       description: 'Number of character engagement actions (like/bookmark)',
     }),
-    wsConnectionsActive: meter.createUpDownCounter('ws.connections.active', {
+    wsConnectionsActive: meter.createUpDownCounter(METRIC_WS_CONNECTIONS_ACTIVE, {
       description: 'Active WebSocket connections',
     }),
-    wsMessagesSent: meter.createCounter('ws.messages.sent', {
+    wsMessagesSent: meter.createCounter(METRIC_WS_MESSAGES_SENT, {
       description: 'Messages sent via WebSocket',
     }),
-    wsMessagesReceived: meter.createCounter('ws.messages.received', {
+    wsMessagesReceived: meter.createCounter(METRIC_WS_MESSAGES_RECEIVED, {
       description: 'Messages received via WebSocket',
     }),
   }
 
   // Revenue metrics
   const revenue: RevenueMetrics = {
-    stripeCheckoutCreated: meter.createCounter('stripe.checkout.created', {
+    stripeCheckoutCreated: meter.createCounter(METRIC_STRIPE_CHECKOUT_CREATED, {
       description: 'Number of Stripe checkout sessions created',
     }),
-    stripeCheckoutCompleted: meter.createCounter('stripe.checkout.completed', {
+    stripeCheckoutCompleted: meter.createCounter(METRIC_STRIPE_CHECKOUT_COMPLETED, {
       description: 'Number of Stripe checkout sessions completed',
     }),
-    stripePaymentFailed: meter.createCounter('stripe.payment.failed', {
+    stripePaymentFailed: meter.createCounter(METRIC_STRIPE_PAYMENT_FAILED, {
       description: 'Number of failed Stripe payments',
     }),
-    stripeSubscriptionEvent: meter.createCounter('stripe.subscription.event', {
+    stripeSubscriptionEvent: meter.createCounter(METRIC_STRIPE_SUBSCRIPTION_EVENT, {
       description: 'Number of Stripe subscription lifecycle events',
     }),
-    stripeEvents: meter.createCounter('stripe.events', {
+    stripeEvents: meter.createCounter(METRIC_STRIPE_EVENTS, {
       description: 'Number of Stripe webhook events processed',
     }),
-    fluxInsufficientBalance: meter.createCounter('flux.insufficient_balance', {
+    fluxInsufficientBalance: meter.createCounter(METRIC_FLUX_INSUFFICIENT_BALANCE, {
       description: 'Number of insufficient flux balance errors',
     }),
   }
 
-  // LLM / Gateway metrics
-  const llm: LlmMetrics = {
-    requestDuration: meter.createHistogram('llm.request.duration', {
-      description: 'LLM gateway request duration in milliseconds',
-      unit: 'ms',
+  // GenAI metrics (semconv: gen_ai.client.*)
+  const genAi: GenAiMetrics = {
+    operationDuration: meter.createHistogram(METRIC_GEN_AI_CLIENT_OPERATION_DURATION, {
+      description: 'GenAI client operation duration',
+      unit: 's',
     }),
-    requestCount: meter.createCounter('llm.request.count', {
-      description: 'Number of LLM gateway requests',
+    operationCount: meter.createCounter(METRIC_GEN_AI_CLIENT_OPERATION_COUNT, {
+      description: 'Number of GenAI client operations',
     }),
-    tokensPrompt: meter.createCounter('llm.tokens.prompt', {
-      description: 'Total prompt tokens consumed',
+    tokenUsageInput: meter.createCounter(METRIC_GEN_AI_CLIENT_TOKEN_USAGE_INPUT, {
+      description: 'Total input (prompt) tokens consumed',
     }),
-    tokensCompletion: meter.createCounter('llm.tokens.completion', {
-      description: 'Total completion tokens consumed',
+    tokenUsageOutput: meter.createCounter(METRIC_GEN_AI_CLIENT_TOKEN_USAGE_OUTPUT, {
+      description: 'Total output (completion) tokens consumed',
     }),
-    fluxConsumed: meter.createCounter('flux.consumed', {
+    fluxConsumed: meter.createCounter(METRIC_FLUX_CONSUMED, {
       description: 'Total flux consumed',
     }),
   }
 
-  // Database metrics
-  const db: DbMetrics = {
-    queryDuration: meter.createHistogram('db.client.operation.duration', {
-      description: 'Database operation duration in milliseconds',
-      unit: 'ms',
-    }),
-    redisCommandDuration: meter.createHistogram('redis.client.command.duration', {
-      description: 'Redis command duration in milliseconds',
-      unit: 'ms',
-    }),
-  }
-
   // Graceful shutdown
   const shutdown = async () => {
     try {
@@ -294,8 +309,7 @@ export function initOtel(env: Env): OtelInstance | undefined {
     auth,
     engagement,
     revenue,
-    llm,
-    db,
+    genAi,
     shutdown,
   }
 }
diff --git a/apps/server/src/middlewares/otel.ts b/apps/server/src/middlewares/otel.ts
index 3c57e8376..4dc6761d9 100644
--- a/apps/server/src/middlewares/otel.ts
+++ b/apps/server/src/middlewares/otel.ts
@@ -39,7 +39,7 @@ export function otelMiddleware(http: HttpMetrics): MiddlewareHandler<HonoEnv> {
         span.setStatus({ code: SpanStatusCode.ERROR, message: `HTTP ${status}` })
       }
 
-      http.requestDuration.record(performance.now() - startTime, {
+      http.requestDuration.record((performance.now() - startTime) / 1000, {
         'http.request.method': method,
         'http.route': path,
         'http.response.status_code': status,
diff --git a/apps/server/src/routes/openai/v1/index.ts b/apps/server/src/routes/openai/v1/index.ts
index b88b06d25..ef5686322 100644
--- a/apps/server/src/routes/openai/v1/index.ts
+++ b/apps/server/src/routes/openai/v1/index.ts
@@ -1,7 +1,7 @@
 import type { Context } from 'hono'
 
 import type { MqService } from '../../../libs/mq'
-import type { LlmMetrics } from '../../../libs/otel'
+import type { GenAiMetrics } from '../../../libs/otel'
 import type { UsageInfo } from '../../../services/billing/billing'
 import type { BillingEvent } from '../../../services/billing/billing-events'
 import type { BillingService } from '../../../services/billing/billing-service'
@@ -69,22 +69,22 @@ function getLlmMetricAttributes(opts: { model: string, type: string, status: num
   }
 }
 
-export function createV1CompletionsRoutes(fluxService: FluxService, billingService: BillingService, configKV: ConfigKVService, billingMq: MqService<BillingEvent>, llm?: LlmMetrics | null) {
+export function createV1CompletionsRoutes(fluxService: FluxService, billingService: BillingService, configKV: ConfigKVService, billingMq: MqService<BillingEvent>, genAi?: GenAiMetrics | null) {
   const logger = useLogger('v1-completions').useGlobalConfig()
   // TODO: Extract this compat route into smaller facades/modules.
   // It currently mixes auth, rate limiting, proxying, billing, telemetry, and event publishing in one transport layer entrypoint.
 
   function recordMetrics(opts: { model: string, status: number, type: string, durationMs: number, fluxConsumed: number, promptTokens?: number, completionTokens?: number }) {
-    if (!llm)
+    if (!genAi)
       return
     const attrs = getLlmMetricAttributes(opts)
-    llm.requestCount.add(1, attrs)
-    llm.requestDuration.record(opts.durationMs, attrs)
-    llm.fluxConsumed.add(opts.fluxConsumed, attrs)
+    genAi.operationCount.add(1, attrs)
+    genAi.operationDuration.record(opts.durationMs / 1000, attrs)
+    genAi.fluxConsumed.add(opts.fluxConsumed, attrs)
     if (opts.promptTokens != null)
-      llm.tokensPrompt.add(opts.promptTokens, attrs)
+      genAi.tokenUsageInput.add(opts.promptTokens, attrs)
     if (opts.completionTokens != null)
-      llm.tokensCompletion.add(opts.completionTokens, attrs)
+      genAi.tokenUsageOutput.add(opts.completionTokens, attrs)
   }
 
   function publishRequestLog(entry: { userId: string, model: string, status: number, durationMs: number, fluxConsumed: number, promptTokens?: number, completionTokens?: number }) {
diff --git a/apps/server/src/utils/observability.ts b/apps/server/src/utils/observability.ts
index 6b4844d8e..109de0201 100644
--- a/apps/server/src/utils/observability.ts
+++ b/apps/server/src/utils/observability.ts
@@ -1,8 +1,14 @@
+// ---------------------------------------------------------------------------
+// Span attribute constants (OTel semantic conventions + AIRI custom)
+// ---------------------------------------------------------------------------
+
+// GenAI semconv attributes — https://opentelemetry.io/docs/specs/semconv/gen-ai/
 export const GEN_AI_ATTR_OPERATION_NAME = 'gen_ai.operation.name'
 export const GEN_AI_ATTR_REQUEST_MODEL = 'gen_ai.request.model'
 export const GEN_AI_ATTR_USAGE_INPUT_TOKENS = 'gen_ai.usage.input_tokens'
 export const GEN_AI_ATTR_USAGE_OUTPUT_TOKENS = 'gen_ai.usage.output_tokens'
 
+// AIRI custom span attributes
 export const AIRI_ATTR_BILLING_FLUX_CONSUMED = 'airi.billing.flux_consumed'
 export const AIRI_ATTR_GEN_AI_INPUT_MESSAGES = 'airi.gen_ai.input.messages'
 export const AIRI_ATTR_GEN_AI_INPUT_TEXT = 'airi.gen_ai.input.text'
@@ -13,9 +19,53 @@ export const AIRI_ATTR_GEN_AI_OUTPUT_TEXT = 'airi.gen_ai.output.text'
 export const AIRI_ATTR_GEN_AI_STREAM = 'airi.gen_ai.stream'
 export const AIRI_ATTR_GEN_AI_STREAM_INTERRUPTED = 'airi.gen_ai.stream_interrupted'
 
+// Server attributes
 export const SERVER_ATTR_ADDRESS = 'server.address'
 export const SERVER_ATTR_PORT = 'server.port'
 
+// ---------------------------------------------------------------------------
+// Metric name constants
+// ---------------------------------------------------------------------------
+
+// HTTP — https://opentelemetry.io/docs/specs/semconv/http/http-metrics/
+export const METRIC_HTTP_SERVER_REQUEST_DURATION = 'http.server.request.duration'
+export const METRIC_HTTP_SERVER_ACTIVE_REQUESTS = 'http.server.active_requests'
+
+// Auth & user (AIRI custom)
+export const METRIC_AUTH_ATTEMPTS = 'auth.attempts'
+export const METRIC_AUTH_FAILURES = 'auth.failures'
+export const METRIC_USER_REGISTERED = 'user.registered'
+export const METRIC_USER_LOGIN = 'user.login'
+export const METRIC_USER_ACTIVE_SESSIONS = 'user.active_sessions'
+
+// Engagement (AIRI custom)
+export const METRIC_CHAT_MESSAGES = 'chat.messages'
+export const METRIC_CHARACTER_CREATED = 'character.created'
+export const METRIC_CHARACTER_DELETED = 'character.deleted'
+export const METRIC_CHARACTER_ENGAGEMENT = 'character.engagement'
+export const METRIC_WS_CONNECTIONS_ACTIVE = 'ws.connections.active'
+export const METRIC_WS_MESSAGES_SENT = 'ws.messages.sent'
+export const METRIC_WS_MESSAGES_RECEIVED = 'ws.messages.received'
+
+// Revenue (AIRI custom)
+export const METRIC_STRIPE_CHECKOUT_CREATED = 'stripe.checkout.created'
+export const METRIC_STRIPE_CHECKOUT_COMPLETED = 'stripe.checkout.completed'
+export const METRIC_STRIPE_PAYMENT_FAILED = 'stripe.payment.failed'
+export const METRIC_STRIPE_SUBSCRIPTION_EVENT = 'stripe.subscription.event'
+export const METRIC_STRIPE_EVENTS = 'stripe.events'
+export const METRIC_FLUX_INSUFFICIENT_BALANCE = 'flux.insufficient_balance'
+
+// GenAI — https://opentelemetry.io/docs/specs/semconv/gen-ai/gen-ai-metrics/
+export const METRIC_GEN_AI_CLIENT_OPERATION_DURATION = 'gen_ai.client.operation.duration'
+export const METRIC_GEN_AI_CLIENT_OPERATION_COUNT = 'gen_ai.client.operation.count'
+export const METRIC_GEN_AI_CLIENT_TOKEN_USAGE_INPUT = 'gen_ai.client.token.usage.input'
+export const METRIC_GEN_AI_CLIENT_TOKEN_USAGE_OUTPUT = 'gen_ai.client.token.usage.output'
+export const METRIC_FLUX_CONSUMED = 'airi.billing.flux.consumed'
+
+// ---------------------------------------------------------------------------
+// Helpers
+// ---------------------------------------------------------------------------
+
 export function getServerConnectionAttributes(baseUrl: string): Record<string, string | number> {
   const url = new URL(baseUrl)
   const attributes: Record<string, string | number> = {