Human Input Revision
This report treats /Users/robertwilhelm/Documents/New project/New_human_input.drawio as the latest human-authored Earthbond ingestion workflow reference.
Recorded author: Johann F.R Wilhelm
Revision log:
docs/operations/HUMAN_INPUT_WORKFLOW_LOG_20260331_REVISION_2.md
Prior reference retained for history:
/Users/robertwilhelm/Documents/New project/human_workflow.drawiodocs/operations/HUMAN_INPUT_WORKFLOW_LOG_20260331.md
What The Revised Human Workflow Now Requires
| Workflow Element | Human Requirement In The Revised Draw.io Source |
|---|---|
| Tenant anchor | core.tenant must act as the upstream tenant anchor and be replicated across clusters. |
| Ingest control | flow_a.ingest_job and flow_a.ingest_job_object define the ingest run and job-to-object link. |
| Object registry | flow_a.object stores the MinIO pointer and object state. |
| State machine | flow_a.object_state_log stores explicit state transitions and changed_by. |
| Metadata extraction | flow_a.object_extract stores CRS, CRS confidence, CRS detection method, epoch, vertical datum, bbox_ecef, and metadata JSON. |
| Typed extract tables | flow_a.object_extract_las and flow_a.object_extract_witsml capture type-specific structured fields. |
| Reproducibility | flow_a.input_snapshot and flow_a.lineage_link capture reproducibility and lineage. |
| Ontology join intent | flow_a.ontology_join_preset encodes workflow-specific ontology join rules. |
| Event-driven handoff | flow_a.outbox emits ObjectCommitted and IngestJobCompleted for PG-B consumption without direct PG-B queries to PG-A. |
| Quarantine | Raw object stays in MinIO while PG-A state changes to QUARANTINED and issue flags persist. |
Where The Current Build Already Matches Strongly
| Revised Workflow Requirement | Current Build Mapping | Status |
|---|---|---|
| Tenant anchor | core.tenants, tenant-scoped project/workflow execution, tenant-aware audit rows |
Matched at concept level |
| Raw bytes in MinIO | MinIO raw-vault storage, staged field-package prefixes, retained raw source objects | Matched |
| Object pointer in PostgreSQL | raw.source_bundles, raw.source_objects |
Matched |
| Immediate audit trail | audit.event_ledger, audit.evidence_packs, audit.integrity_checks |
Matched |
| Metadata capture in PostgreSQL | raw.extracted_fields, raw.source_objects, canonical artifact tables, coverage rows |
Matched in outcome |
| Canonical and ontology mapping added beyond raw capture | ops.canonical_*, semantic.entities, semantic.entity_links |
Implemented |
Where The Current Build Is Partial Or Uses A Different Shape
| Revised Workflow Requirement | Current Build Mapping | Why This Is Partial |
|---|---|---|
flow_a.ingest_job plus flow_a.ingest_job_object |
ops.jobs, ops.upload_sessions, ops.job_results, raw source rows |
The system has ingest job control, but not the exact new flow_a job-to-object contract table pair. |
flow_a.object_state_log |
raw.source_objects state fields plus audit.event_ledger |
State and events exist, but not as one explicit object-state transition table with the exact transition vocabulary and changed_by contract. |
flow_a.object_extract with CRS confidence and detection method |
raw.extracted_fields, metadata profiles, format-specific artifact tables |
Extraction exists, but the exact consolidated table and field set are not implemented one-to-one across all formats. |
flow_a.object_extract_las |
ops.well_logs, ops.canonical_log_artifacts, log interpretation tables |
The build stores richer log semantics, but not in the exact typed extract subtable required by the revised workflow. |
flow_a.object_extract_witsml |
ops.canonical_witsml_wellbores, ops.canonical_witsml_trajectories, ops.canonical_witsml_messages |
WITSML is canonically processed, but not through the exact revised PG-A extract-table shape. |
flow_a.input_snapshot and flow_a.lineage_link |
audit rows, evidence packs, source-link tables, coverage rows | The build preserves provenance and lineage clues, but not yet as the exact snapshot/link tables named in the revised workflow. |
flow_a.ontology_join_preset |
semantic.entities, semantic.entity_links, ontology manuals, canonical join rules in code |
Ontology mapping exists, but not yet as an explicit persisted preset table carrying thresholds and tolerance policies. |
flow_a.outbox and PG-B event consumption |
No exact ingest outbox/event-bus split identified in the live schema | The system publishes artifacts and runs jobs, but it does not yet expose the exact PG-A outbox to PG-B event handoff model defined in the revised workflow. |
| Unified quarantine state semantics | validation gate, blocked states, partial states, retained raw objects, audit issue rows | The behavior exists in parts, but not yet as one strict cross-domain quarantine state machine matching the revised draw.io source. |
What Is Missing As An Exact Implementation Of The Revised Workflow
The build is strong on MinIO plus PostgreSQL provenance, canonical layers, and ontology projection. It is not yet a literal implementation of the new flow_a schema shown in New_human_input.drawio.
- Exact
flow_a.ingest_job_objecttable - Exact
flow_a.object_state_logtransition table - Exact
flow_a.object_extractconsolidated schema with the full revised field set - Exact typed extract subtables for LAS and WITSML in PG-A form
- Exact
flow_a.input_snapshottable - Exact
flow_a.lineage_linktable - Exact
flow_a.ontology_join_presettable - Exact
flow_a.outboxplus event-bus handoff to PG-B without direct query coupling
What The Current Build Added Beyond The Revised Raw-Ingest Workflow
| Added Layer | Current DB Mapping | Why It Exists |
|---|---|---|
| Canonical identity | ops.canonical_wells, ops.canonical_well_aliases, ops.canonical_well_locations | Resolve source inconsistency into stable well truth. |
| Canonical subsurface | ops.canonical_formation_tops, ops.canonical_completion_intervals, ops.canonical_structural_surfaces | Move beyond metadata capture into typed reservoir facts. |
| Canonical operational context | ops.canonical_production_records, ops.canonical_technical_daily_reports, ops.canonical_witsml_* | Provide decision-grade operational context. |
| Ontology projection | semantic.entities, semantic.entity_links | Support ontology reasoning without replacing canonical truth. |
| Decision support | ops.well_reopening_targets, ops.remaining_barrel_estimates | Produce defendable ranking outputs from the canonical layers. |
Conclusion
The revised human workflow is now logged as a distinct authoritative revision.
The current build already matches the MinIO raw vault plus PostgreSQL catalog pattern strongly, and it goes further by implementing canonical and ontology layers. The main gap is that the build still does not implement the revised flow_a tables and event-handshake contract exactly as drawn.