ingestion

Ingesting dependencies from manifests and diagrams involves parsing diagram-based or manifest-based dependency sources, applying filters by priority and type, validating file existence, and deduplicating results to generate structured Read commands for downstream processing.

Yes, ingestion works in both CLI and Trigger.dev contexts. It processes manifest-based and diagram-based dependencies consistently across environments, optionally including deprecated files and enforcing a 25k-token budget for content injection.

Content injection delivers dependencies in chunks to downstream agents as structured, deduplicated Read output suitable for RAG context construction, respecting the 25k-token budget limit for safe content delivery.

The ingestion process automatically validates file existence against the filesystem and deduplicate results, ensuring only valid, unique dependencies are output in the final Read commands for orchestration tools.

Dependency ingestion processes CFN Loop dependency manifests and diagrams using Node.js filesystem operations, child process execution, and path resolution to extract and structure dependencies for orchestration and RAG workflows.