Initial public ModelForge release

2026-09-01 21:30:16 +02:00
commit 7082ab955a
490 changed files with 104252 additions and 0 deletions
@@ -0,0 +1,43 @@
+# Disposable Node Agent identity for the M15 node-loss and re-enrollment rehearsal.
+#
+# Production node credentials are never revoked to prove recovery. This projection runs a second,
+# throwaway Node Agent with its own persisted identity and state volume, so credential loss,
+# revocation and re-enrollment can be exercised end to end without touching GPU Node.
+#
+#   docker compose -p modelforge --env-file .env \
+#     -f docker-compose.node-recovery.yml up -d node-recovery-agent
+services:
+  node-recovery-agent:
+    build:
+      context: .
+      dockerfile: node-agent/Dockerfile
+    hostname: ${MODELFORGE_RECOVERY_AGENT_HOSTNAME:-modelforge-recovery-rehearsal}
+    environment:
+      MODELFORGE_AGENT_CONTROL_PLANE_URL: ${MODELFORGE_RECOVERY_AGENT_URL:-http://api:8000}
+      MODELFORGE_AGENT_ENROLLMENT_TOKEN: ${MODELFORGE_RECOVERY_AGENT_TOKEN:-}
+      MODELFORGE_AGENT_DISPLAY_NAME: ${MODELFORGE_RECOVERY_AGENT_DISPLAY_NAME:-M15 node recovery rehearsal}
+      MODELFORGE_AGENT_IDENTITY_MODE: persisted
+      MODELFORGE_AGENT_IDENTITY_FILE: /data/state/node-id
+      MODELFORGE_AGENT_CREDENTIAL_FILE: /data/state/node-credential
+      MODELFORGE_AGENT_STATE_FILE: /data/state/agent-state.json
+      MODELFORGE_AGENT_TLS_VERIFY: "false"
+      MODELFORGE_AGENT_HEARTBEAT_INTERVAL_SECONDS: "5"
+    volumes:
+      - node-recovery-state:/data/state
+      - node-recovery-cache:/data/hf-cache
+      - node-recovery-artifacts:/data/artifacts
+      - node-recovery-quarantine:/data/quarantine
+    read_only: true
+    tmpfs:
+      - /tmp:size=64m,mode=1777
+    cap_drop:
+      - ALL
+    security_opt:
+      - no-new-privileges:true
+    restart: "no"
+
+volumes:
+  node-recovery-state:
+  node-recovery-cache:
+  node-recovery-artifacts:
+  node-recovery-quarantine: