enrich the setup. setup fixes. minimize the consolidation plan

docs/DEVELOPER_ONBOARDING.md

@@ -95,6 +95,27 @@ StellaOps is a deterministic, offline-first SBOM + VEX platform built as a micro
 - **Disk:** 50 GB free space (for Docker images, volumes, build artifacts)
 - **CPU:** 4 cores minimum, 8 cores recommended
 
+### Runtime Data Assets
+
+Some services need files that are not produced by `dotnet build` or `npm run build`.
+Run this after cloning to get full functionality locally:
+
+```bash
+# ONNX embedding model for semantic search (~80 MB)
+./devops/runtime-assets/acquire.sh --models
+
+# Optional: JDK + Ghidra for binary analysis (~1.6 GB)
+./devops/runtime-assets/acquire.sh --ghidra
+
+# Verify everything
+./devops/runtime-assets/acquire.sh --verify
+```
+
+Without the ONNX model, AdvisoryAI unified search falls back to a reduced-quality
+encoder. All other services work normally. See
+[`devops/runtime-assets/README.md`](../devops/runtime-assets/README.md) for the
+full asset inventory and Docker volume mount options.
+
 ---
 
 ## Quick Start