RETL File Destination

condor-retl-file is a first-party destination connector that writes RETL State Operations and Event imports to local CSV export drops. It exposes the retl/file connector through the retl.destinations entry point group.

Surfaces

• csv_state_operations accepts State Operations, supports upsert and remove, and writes upserts.csv and removes.csv when those operation kinds are present.
• csv_event_imports accepts Event imports and writes imports.csv.

Both surfaces accept RETL's common identifier types without requiring a specific identifier, because the connector writes identifier values through to the CSV rather than translating them for a partner API.

Each non-dry-run submission creates a new timestamped export directory under output_dir and writes a manifest.json with file names, row counts, byte sizes, SHA-256 checksums, RETL batch ids, and operation counts. Successful file writes produce definitive succeeded delivery evidence.

Binding Config

The file connector does not use secrets or credentials.

• output_dir is required and points to the parent directory for export drops.
• file_batch_max_rows controls RETL file batch planning. It defaults to 10000.
• create_parent_dirs controls whether missing output parents are created during non-dry-run submission. It defaults to true.

output_dir, file_batch_max_rows, and create_parent_dirs are namespace-loadable public config:

destination = retl.destinations.load(
    "retl/file",
    binding_name="file_exports",
    config_namespace="destinations.file",
)

CSV Shape

CSV files use stable structured columns:

• operation
• record_identity
• identifiers_json
• key_json
• payload_json
• target
• occurred_at
• collect_id
• sequence_order
• payload_fingerprint

V1 writes new export drops for every submission. It does not mutate existing CSV files and does not claim exactly-once file delivery; downstream consumers should use manifest batch ids and checksums for deduplication.