AIBOM Configuration

The AIBOM module ingests pre-generated Cisco AI BOM JSON reports and maps them onto container images already present in Cartography.

Cartography does not run the scanner in this module. It only ingests JSON artifacts from local disk or S3.

Why this module exists

Traditional image inventory tells you what packages and vulnerabilities exist in a container. It does not tell you whether that container includes AI agents, models, prompts, tools, memory layers, or other agentic building blocks.

This module adds that missing inventory layer and ties it to the production graph through ECRImage, so you can ask questions such as:

• Which production images contain AI agents?

• Which agents use tools, prompts, models, or memory?

• Which scans failed or did not match a known image in the graph?

Input format

Each JSON file must be an envelope wrapping the native scanner output with the image URI that should be matched to the graph.

{
  "image_uri": "000000000000.dkr.ecr.us-east-1.amazonaws.com/example-repository:v1.0",
  "scan_scope": "/app",
  "scanner": {
    "name": "cisco-aibom",
    "version": "0.4.0"
  },
  "report": {
    "aibom_analysis": {
      "...": "native scanner output"
    }
  }
}

Field	Required	Description
image_uri	Yes	Image URI to map into the graph. Tag-based and digest-based URIs are both supported.
report	Yes	Wrapper object containing the native aibom_analysis.
scan_scope	No	Path or scope scanned inside the image or extracted filesystem. Stored on AIBOMSource.
scanner.name	No	Scanner name. Defaults to cisco-aibom.
scanner.version	No	Scanner version. Falls back to aibom_analysis.metadata.analyzer_version.

The native source payload may optionally include:

• workflows

• relationships

• source_kind

• category-specific component fields such as model_name, framework, and label

The module preserves those optional fields when present.

ECR-first linking behavior

AIBOM resolves each report to the most canonical ECRImage already in the graph:

1. Prefer ECRImage nodes where type = "manifest_list".

2. Fall back to ECRImage nodes where type = "image" only when no manifest list exists for the tagged image.

This avoids duplicating detections across platform-specific child images while still supporting single-platform images.

Provenance behavior

Cartography now preserves source provenance even when component inventory is not loaded:

• If a source has a non-completed status, Cartography loads AIBOMSource but skips components, workflows, and relationships.

• If image_uri does not resolve to an ECRImage, Cartography still loads AIBOMSource with image_matched = false for troubleshooting.

This makes stale coverage, failed scans, and mismatched image URIs visible in the graph instead of silently disappearing.

Prerequisite

Run ECR ingestion before AIBOM ingestion so ECRRepositoryImage and ECRImage nodes exist. In the default sync order AIBOM runs after AWS automatically.

Results layout

The AIBOM module ingests every *.json file under the configured directory or S3 prefix as part of a single snapshot. Keep only the latest scan per image in the results location. If older reports for the same image are also present, their scans and detections will all be loaded in that snapshot because they share the same update_tag.

Run with local files

cartography \
  --selected-modules aibom \
  --aibom-results-dir /path/to/aibom-results

Run with S3

cartography \
  --selected-modules aibom \
  --aibom-s3-bucket my-aibom-bucket \
  --aibom-s3-prefix reports/

--aibom-s3-prefix is optional and defaults to an empty prefix.

Observability counters

• aibom_reports_processed

• aibom_sources_total

• aibom_sources_matched

• aibom_sources_unmatched

• aibom_sources_skipped_incomplete

• aibom_components_loaded_<category>

• aibom_relationships_loaded_<relationship_type>