Diagram generation: ref-arch-driven procedure + spec validator + KB enrichment

The diagram path now follows a documented standard procedure (lookup
the closest Oracle Architecture Center reference → confirm components
→ author absolute_layout → spec validator → render → visually verify)
and ships persistent guardrails so layout regressions can't recur.

Persistent procedure changes (apply to all users, all sessions):
- tools/diagram_spec_validator.py — geometry checks (CONTAINER_TOO_THIN,
  CONTAINER_PADDING_VIOLATION, LABEL_OVERFLOW_PARENT) run BEFORE either
  renderer (drawio + PPTX). Catches the subnet-collapse / label-overflow
  bugs that the post-render drawio validator missed.
- tools/oci_diagram_gen.py + tools/oci_pptx_diagram_gen.py — call the
  spec validator before emitting any output. Adds mysql / mysql_heatwave
  type aliases.
- tools/archcenter_pattern_lookup.py — scores against cached page
  descriptions (not just the 1-line summary), supports --queries for
  multi-fragment composition, and applies synonym expansion via
  kb/architecture-center/synonyms.yaml so "LB HA cross AD" matches
  "load balancer high availability availability domain".
- kb/architecture-center/synonyms.yaml — canonical synonym table
  (load balancer, autonomous database, data guard, …) used by the
  lookup scorer.

KB enrichment:
- tools/archcenter_description_fetcher.py + 121 cached _description.md
  under kb/diagram/assets/archcenter-refs/<slug>/. Removes the runtime
  dependency on docs.oracle.com when authoring specs and feeds the
  pattern-lookup scorer.
- 110+ cached .drawio / .svg / .png references for offline reuse,
  plus the OCI Toolkit v24.2 import (kb/diagram/assets/oci-toolkit-drawio).

Documentation:
- docs/skill/output-formats.md — new "Standard diagram-generation
  procedure (MANDATORY)" + geometry rules + the new validator entry.
- SKILL.md option 2 — references the mandatory procedure.
- README.md — describes the spec validator, archcenter_pattern_lookup
  and description fetcher, and updates the KB-health table.

Tooling that backs the procedure (cumulative across recent sessions):
tools/archcenter_case_runner.py, archcenter_batch_driver.py,
archcenter_zip_downloader.py, drawio_visual_validator.py,
drawio_fidelity_eval.py, harvest_drawio_icon.py, import_oci_library.py,
oci_pptx_diagram_gen.py, oci_pptx_render.py, refresh_pptx_icon_index.py.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

kb/diagram/assets/archcenter-refs/fraud-detection-nvidia-morpheus-compute-cloud/_description.md

@@ -0,0 +1,373 @@
+# Stream fraud detection with NVIDIA Morpheus on Oracle Compute Cloud@Customer
+
+- Source: https://docs.oracle.com/en/solutions/fraud-detection-nvidia-morpheus-compute-cloud/index.html
+- Date: 2025-08
+- Type: built-deployed
+- Services: compute
+- Tags: ai-ml, security
+
+## Summary (catalog)
+
+Real-time fraud detection with NVIDIA Morpheus on Compute Cloud@Customer. GPU-accelerated ML pipeline for transaction monitoring. On-premises deployment for data sovereignty requirements.
+
+## Architecture (fetched from source)
+
+Learn About Streaming Fraud Detection 
+ 
+ 
+ 
+ 
+- 
+ 
+- 
+ 
+ 
+ 
+ 
+ 
+- 
+ 
+ 
+ 
+ 
+ 
+ 
+ 
+ 
+ 
+- 
+ 
+ 
+ 
+ 
+ 
+ 
+ 
+ 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+- 
+ 
+ 
+ 
+ 
+
+ Previous 
+ Next 
+ JavaScript must be enabled to correctly display this content
+ 
+ 
+
+ 
+
+ 
+ 
+ 
+Learn About Streaming Fraud
+ Detection
+
+ 
+
+ 
+
+ 
+ 
+ Oracle Compute Cloud@Customer , a key component of the Oracle Roving Edge Infrastructure portfolio, provides organizations a scalable solution to process sensitive data
+ securely, with low latency—close to its source. 
+ 
+
+ 
+ 
+In this solution playbook, you learn how to use the NVIDIA Morpheus cybersecurity framework to deploy real-time, AI-driven fraud detection using a
+ GPU-accelerated Morpheus pipeline, and publish the results using Compute Cloud@Customer . The solution enables instant fraud detection without waiting for batch jobs and
+ keeps your data secure by locally processing it at the edge.
+ 
+
+ 
+ 
+
+ 
+ 
+Before You Begin
+
+ 
+
+ 
+
+ Ensure you perform the deployment on a host machine with the
+ following environment settings: 
+ 
+ 
+
+ 
+ 
+- Operating System : Ubuntu 24.04 LTS
+ 
+ 
+- Platform : NVIDIA AI Enterprise on a single-node Oracle Compute Cloud@Customer instance equipped with an NVIDIA L40S GPU.
+ 
+ 
+ 
+Before you begin, ensure the following tools are installed on the host machine:
+
+ 
+ 
+- Docker and Docker Compose ( docker compose V2 )
+ 
+ 
+- Git and Git LFS ( git-lfs )
+ 
+ 
+- Python 3 and Pip 
+ 
+ 
+
+ 
+
+ 
+ 
+Workflow
+
+ 
+
+ 
+The workflow leverages the NVIDIA Morpheus cybersecurity framework to perform GPU-accelerated inference on a stream of financial
+ transaction data. 
+ 
+
+ 
+The following diagram shows a workflow through three swimlanes; Host
+ environment, Apache Kafka , and the Morpheus pipeline:
+ 
+
+ 
+ Description of the illustration ai-driven-fraud-detection-workflow-arch.png 
+
+The pipeline ingests live transaction data through Apache Kafka , performs graph-based contextual analysis using a pre-trained Graph Sample and Aggregate ( GraphSAGE ) model, and executes final fraud classification using XGBoost . All stages are accelerated using NVIDIA 
+ RAPIDS libraries (cuDF, cuML), making the entire workflow GPU-optimized for high throughput.
+ The following is how the process flows:
+ 
+
+ 
+ 
+- Transaction data (.csv) is produced by a Python producer.
+ 
+ 
+- Data goes into Kafka topic INPUT stream.
+ 
+ 
+- Data is processed through the Morpheus pipeline in steps: reading from Kafka source, deserializing, constructing a graph, interfacing with Graph Neural Network (GNN) interface ( GraphSAGE ), classifying with
+ XGBoost , serializing results, and writing to Kafka sink.
+ 
+ 
+- Results are sent to Kafka topic OUTPUT stream.
+ 
+ 
+- A Python consumer receives output and provides live fraud prediction.
+ 
+ 
+ 
+ Model Provenance 
+
+ 
+The inference pipeline at the core of this architecture uses two pre-trained
+ machine learning models: a GraphSAGE GNN and an XGBoost classifier. These models were generated using a separate training pipeline, which is
+ included in the Morpheus repository for reference.
+ 
+
+ 
+ 
+- Training Script Location: 
+ examples/gnn_fraud_detection_pipeline/training.py .
+ 
+ 
+- Process: The script processes a labeled historical dataset to
+ train the GNN on graph-based features and the XGBoost model on
+ the resulting embeddings.
+ 
+ 
+ 
+In this solution, you don't have to run the training script becausethe
+ pre-trained models are already provided. Focus on deploying and running the real-time
+ inference pipeline.
+
+ 
+This architecture supports the following components:
+
+ 
+ 
+- Oracle Compute Cloud@Customer 
+ Oracle Compute Cloud@Customer is fully-managed, rack-scale infrastructure
+ that lets you use OCI Compute anywhere. Gain the benefits of cloud automation
+ and economics in your data center by running OCI Compute and GPU shapes with storage and networking
+ services on Compute Cloud@Customer . You can run applications and harness the power
+ of GenAI on cloud infrastructure in your data
+ center while helping address data residency,
+ security, and low-latency connections to local
+ resources and real-time operations.
+ 
+
+ 
+ 
+- RAPIDS 
+ cuDF/cuML 
+ RAPIDS cuDF/cuML are a suite of GPU-accelerated libraries for high-performance data
+ manipulation and machine learning utilities within the Morpheus
+ pipeline.
+ 
+
+ 
+ 
+- Docker + Conda 
+ Docker + Conda provide a layered approach to dependency management, using Docker for OS-level isolation and Conda for managing the complex Python environment inside the container.
+ 
+
+ 
+ 
+ 
+
+ 
+
+ 
+ 
+Considerations for
+ Production
+
+ 
+
+ 
+When implementing this solution in a production environment, consider the
+ scalability and resilience that Kubernetes can provide. You can migrate this solution to OCI Kubernetes Engine ( OKE ) by:
+ 
+
+ 
+ 
+- Containerizing the producer and consumer helper scripts. 
+ 
+- Deploying Kafka using a production-grade Kubernetes operator.
+ 
+ 
+- Deploying the Morpheus pipeline as a Kubernetes job or deployment.
+ 
+ 
+ 
+
+ 
+
+ 
+ 
+About Required Services and
+ Roles
+
+ 
+
+ 
+This solution requires the following services and roles:
+
+ 
+ 
+- Oracle Compute Cloud@Customer 
+ 
+- NVIDIA AI Enterprise 6.0
+ 
+ 
+- 
+ 
+Ubuntu Linux (or a compatible Linux distribution)
+
+ 
+ 
+- Docker 
+ 
+- NVIDIA Morpheus 25.02
+ 
+ 
+ 
+These are the roles needed for each service.
+
+ 
+
+ 
+ 
+ 
+ Service Name: Role 
+ Required to... 
+ 
+ 
+ 
+ 
+ Oracle Compute Cloud@Customer : administrator 
+ Configure and deploy the NVIDIA AI Enterprise virtual machine instance, manage network resources, and ensure access
+ to the NVIDIA L40S GPU.
+ 
+ 
+ 
+ Ubuntu Linux: root or user with sudo 
+ privileges
+ 
+ Install prerequisite software ( Docker , Git), manage system services, and execute Docker commands.
+ 
+ 
+ 
+ NVIDIA AI Enterprise : Account User
+ 
+ Pull the required NVIDIA Morpheus container image from the NVIDIA GPU Cloud ( NGC ) catalog.
+ 
+ 
+ 
+ 
+ 
+
+ 
+See Oracle Products, Solutions, and Services to get what you need.
+ 
+
+ 
+
+ 
+
+ 
+
+ 
+
+ 
+
+ 
+ Title and Copyright Information 
+
+ 
+
+ 
+Fraud detection with NVIDIA Morpheus on Compute Cloud@Customer and Private Cloud Appliance
+
+ 
+G38649-02
+
+ 
+August 2025
+
+ 
+ Copyright © 2025, 
+
+Oracle and/or its affiliates.