Document Batch (GCS)
Category: Healthcare Standards
Node id: document-batch
Standards: PHI, IEC-62304-Class-C
Process every PDF under a GCS bucket/prefix: extract → map to FHIR R4 → upsert via transaction, with per-file human-review gating. Concurrency-bounded; one file's failure never aborts the batch.
What it does
- Batch PDF ingestion from Google Cloud Storage
- Concurrency-bounded, per-file isolation (one failure never aborts the batch)
- Per-file human-review gating of low-confidence extractions
- Idempotent FHIR upsert (document_id = object name)
Runtime
- Performance tier:
variable
Input ports
| Name | Type | Required | Description |
|---|---|---|---|
gcs_bucket | text | No | GCS bucket (overrides config) |
prefix | text | No | Object prefix to scope the batch (overrides config) |
Output ports
| Name | Type | Required | Description |
|---|---|---|---|
summary | json | No | Counts: total, written, needs_review, failed |
results | json | No | Per-file results (object, document_type, written, error, ...) |
review_queue | json | No | Object names flagged needs_review (not written) |
success | boolean | No | True when no file failed |
Settings
Basic
| Setting | Name | Type | Required | Default | Allowed values | Description |
|---|---|---|---|---|---|---|
| GCS Bucket | gcs_bucket | string | Yes | Bucket containing the PDFs | ||
| Prefix | prefix | string | No | Object prefix to scope the batch | ||
| Suffix Filter | suffix | string | No | .pdf | Only process objects ending with this (case-insensitive) | |
| Max Files | max_files | number | No | 0 | min 0 | Cap the number of files processed (0 = all) |
| Model | model | string | No | gemini-2.5-pro | Gemini model (falls back to GEMINI_MODEL) |
Fhir
| Setting | Name | Type | Required | Default | Allowed values | Description |
|---|---|---|---|---|---|---|
| FHIR URL | fhir_url | string | No | Target FHIR server base URL (required unless dry_run) | ||
| FHIR Auth Token | auth_token | secret | No | Bearer token for the FHIR server | ||
| MRN Identifier System | identifier_system | string | No | urn:tietai:hydra:mrn | ||
| Document Identifier System | document_identifier_system | string | No | urn:tietai:hydra:document |
Auth
| Setting | Name | Type | Required | Default | Allowed values | Description |
|---|---|---|---|---|---|---|
| GCP Project | gcp_project | string | No | Vertex AI project (falls back to GEMINI_PROJECT / GOOGLE_CLOUD_PROJECT) | ||
| GCP Location | gcp_location | string | No | global | Vertex AI location (falls back to GEMINI_LOCATION) |
Advanced
| Setting | Name | Type | Required | Default | Allowed values | Description |
|---|---|---|---|---|---|---|
| Max Concurrency | max_concurrency | number | No | 4 | min 1 | Number of files processed in parallel |
| Dry Run | dry_run | boolean | No | false | Extract + map but do not write to FHIR | |
| Write When Needs Review | write_when_needs_review | boolean | No | false | Write even low-confidence documents (default: route to review_queue) | |
| Min Confidence | min_confidence | number | No | 0.6 | min 0; max 1 | Below this, a document is routed to review instead of written |
Runtime notes
- Auth via GCP Application Default Credentials / workload identity (GCS + Vertex).
- Schedule this node's pipeline for recurring bucket processing.
- Set dry_run to validate extraction/mapping without writing to FHIR.
See Build any workflow for how to put nodes together and Ports, types and data flow for how port types connect.
Something unclear? Send documentation feedback by email.