Ingest artifacts

Scan output is versioned under each scanId. On AWS the S3 prefix is:

helix/flowgraphs/{scanId}/

Locally: scans/{scanId}/.

Layout (summary)

{scanId}/
  manifest.json
  reports/
    technical-graph.json      # merged graph (query source of truth)
    feature-resolution.json
    fe-be-bridge.jsonl
  repos/{repoKey}/
    files.jsonl
    technical-graph-partial.json
    …
  ai/
    feature-display-titles.json
    feature-insights/{featureId}.json
    endpoint-qa/{endpointId}.json
    …

Scan pipeline (AWS)

InitScan → Map(AST per repo, Fargate) → MergeGraph (Lambda) → Map(AI per repo, Fargate) → MarkReady

Progress: workers → DynamoDB + AppSync publishScanProgress → UI subscription.

Incremental scans

Per-file fingerprints allow re-parsing only changed paths on new commits. Shared barrel or parser version bumps may still require full repo rescans.

Neptune

Merge step optionally upserts openCypher into Neptune (:TechnicalNode, [:REL]). Query Lambda currently reads S3 JSON; Neptune is optional for traversals and laptop debugging via SSM tunnel.

CLI

npm run ingest -- --roots ~/code/igentify/ng-access-service ...

See repo docs/Ingest-artifacts.md for the full contract (manifest version, JSONL record types).