Elysia Tools
Navigation
Workflow Playbook
Turn mixed workbook and CSV sources into traceable, schema-aware outputs for SQL loading, Parquet pipelines, and object-storage batch handoff.
Hubs
An ETL run is easier to audit when every input has a stable source ID and batch ID. Record the file or API capture, object-storage prefix when relevant, workbook sheet or range, and the expected row and field shape. Keep raw inputs immutable so a normalization fix can be replayed instead of guessed from a changed file.
Land API pages or storage batches first, then resolve CSV encoding and delimiters. Use the sheet splitter and range extractor to avoid mixing notes or totals with data. The column mapper establishes names and order; the unpivot normalizer converts month-as-column layouts into records when the downstream table needs a long shape. Merge only compatible workbooks, and make the incremental cursor part of the lineage rather than an invisible setting.
The JSON Schema tool provides a candidate contract from headers and examples. Compare it with the migration plan, the business key, nullability, and the selected database dialect before accepting generated INSERT statements. The tools prepare text and structure; they do not replace a database transaction or a human decision about type changes.
Before export, reconcile input rows with accepted, duplicate, rejected, and output rows. Check the final field set, types, key uniqueness, and any schema drift. Parquet and NDJSON suit an analytics handoff, while a CSV pack with a manifest makes sheet-level review easier. If a check fails, quarantine the affected branch and retain its lineage instead of publishing a partial result as complete.
Workflow playbook
Bring paginated API records into a sheet when needed, process the approved object-storage batch, and detect CSV delimiter or encoding before batch conversion. Assign source and batch identifiers and preserve the raw-to-landed relationship before changing field values.
Separate multi-sheet workbooks or extract only the approved ranges, then rename, reorder, or remove columns. Unpivot wide period columns when the destination is row-oriented, and flatten or restore nested JSON only when that shape is explicitly part of the target contract.
Merge workbooks only when their sheet and column contract is compatible. For recurring loads, append new rows with a reviewed time or primary-key cursor; compare source, retained, duplicate, and rejected counts and keep the cursor with the batch lineage for safe reruns.
Use API To Sheet for paginated JSON that must first become a workbook, S3 Batch Processor for an approved storage prefix, and the CSV detection or batch tools when delimiter and encoding vary. Record which branch produced each landed artifact.
Split workbooks or extract named ranges when sheets contain separate datasets; map columns before merging; unpivot monthly or category columns when the target expects one observation per row; use the JSON transformer only when nested structure is part of the contract.
Treat inferred JSON Schema as a reviewable proposal, use the migration planner for table and column planning, and generate SQL INSERT statements only after types, keys, quoting, and the target dialect have been checked. Choose Parquet or NDJSON for pipeline handoff and a CSV pack when sheet-level inspection is required.
Infer a candidate JSON Schema from representative rows, turn the approved fields into a database migration plan, and generate SQL INSERT statements for the selected dialect. Inspect dates, numbers, nulls, identifiers, escaping, and target-column order before executing anything outside this workflow.
Export the accepted table shape to Parquet or NDJSON for an analytics pipeline, or package each sheet as CSV with a manifest for inspection and transfer. Attach row and field checks plus lineage metadata to the handoff; the receiving system still owns upload, permissions, and final commit.