update

deployments/master/docker-compose.yml

@@ -23,7 +23,7 @@ services:
     - ALL
   neuroforge:
     image: git.send.nrw/sendnrw/glpi-neuroforge-mega-neuroforge:${IMAGE_TAG:?Set IMAGE_TAG
-      to an immutable release tag, for example 1.6.0}
+      to an immutable release tag, for example 1.6.2}
     command:
     - -data
     - /app/data
@@ -57,14 +57,15 @@ services:
       NEUROFORGE_WORKER_DEFAULT_MAX_ATTEMPTS: ${NEUROFORGE_WORKER_DEFAULT_MAX_ATTEMPTS:-3}
       NEUROFORGE_WORKER_RETRY_BACKOFF_SECONDS: ${NEUROFORGE_WORKER_RETRY_BACKOFF_SECONDS:-15}
       NEUROFORGE_WORKER_MAX_QUEUED_JOBS: ${NEUROFORGE_WORKER_MAX_QUEUED_JOBS:-5000}
+      NEUROFORGE_WORKER_MAX_QUEUED_PAYLOAD_MB: ${NEUROFORGE_WORKER_MAX_QUEUED_PAYLOAD_MB:-128}
       NEUROFORGE_WORKER_MASTER_APPLY_MAX_ATTEMPTS: ${NEUROFORGE_WORKER_MASTER_APPLY_MAX_ATTEMPTS:-5}
       NEUROFORGE_WORKER_MASTER_APPLY_BACKOFF_SECONDS: ${NEUROFORGE_WORKER_MASTER_APPLY_BACKOFF_SECONDS:-5}
-      NEUROFORGE_WORKER_JOB_RETENTION_HOURS: ${NEUROFORGE_WORKER_JOB_RETENTION_HOURS:-168}
-      NEUROFORGE_WORKER_MAX_TERMINAL_JOBS: ${NEUROFORGE_WORKER_MAX_TERMINAL_JOBS:-20000}
+      NEUROFORGE_WORKER_JOB_RETENTION_HOURS: ${NEUROFORGE_WORKER_JOB_RETENTION_HOURS:-24}
+      NEUROFORGE_WORKER_MAX_TERMINAL_JOBS: ${NEUROFORGE_WORKER_MAX_TERMINAL_JOBS:-2000}
       NEUROFORGE_GRAPH_BACKFILL_ENABLED: ${NEUROFORGE_GRAPH_BACKFILL_ENABLED:-true}
       NEUROFORGE_GRAPH_BACKFILL_INTERVAL_SECONDS: ${NEUROFORGE_GRAPH_BACKFILL_INTERVAL_SECONDS:-10}
-      NEUROFORGE_GRAPH_BACKFILL_BATCH_SIZE: ${NEUROFORGE_GRAPH_BACKFILL_BATCH_SIZE:-64}
-      NEUROFORGE_GRAPH_BACKFILL_MAX_QUEUED: ${NEUROFORGE_GRAPH_BACKFILL_MAX_QUEUED:-256}
+      NEUROFORGE_GRAPH_BACKFILL_BATCH_SIZE: ${NEUROFORGE_GRAPH_BACKFILL_BATCH_SIZE:-16}
+      NEUROFORGE_GRAPH_BACKFILL_MAX_QUEUED: ${NEUROFORGE_GRAPH_BACKFILL_MAX_QUEUED:-64}
       NEUROFORGE_GRAPH_BACKFILL_MIN_DEGREE: ${NEUROFORGE_GRAPH_BACKFILL_MIN_DEGREE:-3}
       NEUROFORGE_GRAPH_CANDIDATE_MULTIPLIER: ${NEUROFORGE_GRAPH_CANDIDATE_MULTIPLIER:-6}
       NEUROFORGE_GRAPH_RETRY_AFTER_MINUTES: ${NEUROFORGE_GRAPH_RETRY_AFTER_MINUTES:-360}
@@ -137,7 +138,7 @@ services:
     stop_grace_period: 35s
   agent-data-init:
     image: git.send.nrw/sendnrw/glpi-neuroforge-mega-agent-data-init:${IMAGE_TAG:?Set
-      IMAGE_TAG to an immutable release tag, for example 1.6.0}
+      IMAGE_TAG to an immutable release tag, for example 1.6.2}
     restart: 'no'
     user: 0:0
     volumes:
@@ -151,7 +152,7 @@ services:
     - FOWNER
   agent:
     image: git.send.nrw/sendnrw/glpi-neuroforge-mega-agent:${IMAGE_TAG:?Set IMAGE_TAG
-      to an immutable release tag, for example 1.6.0}
+      to an immutable release tag, for example 1.6.2}
     restart: unless-stopped
     environment:
       AI_CONTENT_LABEL_ENABLED: ${AI_CONTENT_LABEL_ENABLED:-}
@@ -373,7 +374,7 @@ services:
     stop_grace_period: 20s
   knowledge:
     image: git.send.nrw/sendnrw/glpi-neuroforge-mega-knowledge:${IMAGE_TAG:?Set IMAGE_TAG
-      to an immutable release tag, for example 1.6.0}
+      to an immutable release tag, for example 1.6.2}
     restart: unless-stopped
     environment:
       APP_MODE: ${KB_APP_MODE:-editor}
@@ -414,7 +415,7 @@ services:
     stop_grace_period: 35s
   control:
     image: git.send.nrw/sendnrw/glpi-neuroforge-mega-control:${IMAGE_TAG:?Set IMAGE_TAG
-      to an immutable release tag, for example 1.6.0}
+      to an immutable release tag, for example 1.6.2}
     restart: unless-stopped
     environment:
       CONTROL_ADDR: :8070
@@ -543,7 +544,7 @@ services:
     - prometheus-secrets:/etc/prometheus-secrets:ro
     depends_on:
       neuroforge:
-        condition: service_healthy
+        condition: service_started
       prometheus-secrets-init:
         condition: service_completed_successfully
       prometheus-config-init:

deployments/master/monitoring/grafana/dashboards/neuroforge-master-subagents.json

@@ -9,7 +9,7 @@
   ],
   "timezone": "browser",
   "schemaVersion": 41,
-  "version": 1,
+  "version": 2,
   "refresh": "15s",
   "time": {
     "from": "now-6h",
@@ -1443,6 +1443,45 @@
           "sort": "desc"
         }
       }
+    },
+    {
+      "id": 26,
+      "title": "Durable Job Payload Bytes",
+      "type": "timeseries",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prometheus"
+      },
+      "gridPos": {
+        "x": 0,
+        "y": 56,
+        "w": 24,
+        "h": 8
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes"
+        },
+        "overrides": []
+      },
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": true
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "none"
+        }
+      },
+      "targets": [
+        {
+          "expr": "neuroforge_job_payload_bytes",
+          "legendFormat": "{{state}}",
+          "refId": "A"
+        }
+      ]
     }
   ]
 }

deployments/master/monitoring/grafana/provisioning/dashboards/neuroforge.yml

@@ -0,0 +1,12 @@
+apiVersion: 1
+
+providers:
+  - name: NeuroForge
+    orgId: 1
+    folder: NeuroForge
+    type: file
+    disableDeletion: true
+    updateIntervalSeconds: 30
+    allowUiUpdates: false
+    options:
+      path: /var/lib/grafana/dashboards

deployments/master/monitoring/grafana/provisioning/datasources/prometheus.yml

@@ -1,4 +1,5 @@
 apiVersion: 1
+
 datasources:
   - name: Prometheus
     uid: prometheus

deployments/master/monitoring/prometheus/alerts.yml

@@ -56,3 +56,19 @@ groups:
         annotations:
           summary: "Knowledge graph remains highly isolated"
           description: "More than 80% of memories are still isolated after 30 minutes."
+
+      - alert: NeuroForgePendingPayloadHigh
+        expr: neuroforge_job_payload_bytes{state="pending"} > 104857600
+        for: 5m
+        labels: {severity: warning}
+        annotations:
+          summary: "NeuroForge pending job payload memory is high"
+          description: "Pending durable job payload/result bytes exceed 100 MiB for 5 minutes; backfill should remain below the v1.6.2 safety budget."
+
+      - alert: NeuroForgeTerminalPayloadHigh
+        expr: neuroforge_job_payload_bytes{state="terminal"} > 67108864
+        for: 10m
+        labels: {severity: warning}
+        annotations:
+          summary: "NeuroForge terminal job payload retention is high"
+          description: "Terminal job payload/result bytes exceed 64 MiB. Completed vector.relink jobs should be compacted automatically in v1.6.2."

deployments/master/preflight.sh

@@ -1,15 +1,14 @@
 #!/usr/bin/env sh
 set -eu
-bad=0
-check() { key="$1"; val=$(grep -E "^${key}=" .env | head -1 | cut -d= -f2- || true); if [ -z "$val" ] || echo "$val" | grep -Eq 'CHANGE_ME|example\.invalid|192\.0\.2\.'; then echo "SETZEN: $key"; bad=1; fi; }
-for k in GLPI_URL GLPI_CLIENT_ID GLPI_CLIENT_SECRET GLPI_USERNAME GLPI_PASSWORD OLLAMA_BASE_URL OLLAMA_URLS; do check "$k"; done
-if [ "$bad" -ne 0 ]; then echo "Preflight fehlgeschlagen: Platzhalter ersetzen."; exit 1; fi
-command -v docker >/dev/null 2>&1 || { echo "FEHLT: docker"; exit 1; }
-docker compose version >/dev/null
+[ -f .env ] || { echo 'missing .env'; exit 1; }
+required='IMAGE_TAG NEUROFORGE_ADMIN_TOKEN NEUROFORGE_APP_API_KEY NEUROFORGE_INTEGRATION_TOKEN NEUROFORGE_CONTROL_READ_TOKEN NEUROFORGE_WORKER_TOKEN NEUROFORGE_METRICS_TOKEN KB_INTEGRATION_TOKEN CONTROL_READ_TOKEN GLPI_URL GLPI_CLIENT_ID GLPI_CLIENT_SECRET GLPI_USERNAME GLPI_PASSWORD GRAFANA_ADMIN_PASSWORD'
+for k in $required; do
+  v=$(awk -F= -v key="$k" '$1==key {sub(/^[^=]*=/,""); print; exit}' .env)
+  [ -n "$v" ] || { echo "missing: $k"; exit 1; }
+done
+if grep -Eq '(^|=)(https?://)?192\.0\.2\.|example\.invalid|CHANGE_ME|YOUR-' .env; then
+  echo 'placeholder/example values remain in .env'; exit 1
+fi
+[ "$(awk -F= '$1=="IMAGE_TAG"{print $2}' .env)" = "1.6.2" ] || { echo 'IMAGE_TAG must be 1.6.2'; exit 1; }
 docker compose --profile research --profile monitoring config >/dev/null
-[ -f monitoring/prometheus/prometheus.yml.template ]
-[ -f monitoring/prometheus/alerts.yml ]
-[ -f monitoring/grafana/dashboards/neuroforge-master-subagents.json ]
-mkdir -p knowledge staging backups
-./monitoring/validate.sh
-echo "Master preflight: OK"
+echo 'master preflight: OK'