Release Notes — v1.7

Release date: March 2026

This release adds exports — declarative secondary output destinations that write thread data to additional locations in configurable formats, without duplicating pipeline logic.

Also includes enhanced flow SVG visualization with validation, quarantine, assertions, and export nodes, plus a new Sankey-style waterfall replacing the horizontal bar chart.

---

Exports (Secondary Outputs)

Configuration

Declare exports at the thread, weave, or loom level. Each export writes the same post-mapping, audit-injected DataFrame as the primary target.

exports:
  - name: parquet_archive
    description: "Daily archive for external consumers"
    type: parquet
    path: /lakehouse/archive/orders/${run.timestamp}/
    partition_by: [region]
    on_failure: warn

  - name: csv_extract
    type: csv
    path: /lakehouse/exports/orders_latest/
    options:
      header: "true"
      delimiter: "|"

Supported Formats

Five Fabric-native Spark formats are supported:

Format	Path	Alias	Notes
delta	yes	yes	Metastore or OneLake path
parquet	yes	no	Compression via options
csv	yes	no	Header, delimiter via options
json	yes	no	One object per line
orc	yes	no	Hive-compatible columnar

Cascade Inheritance

Exports cascade additively through loom → weave → thread, following the same pattern as audit_columns:

• All levels contribute exports; collected by name
• Same-named exports at a lower level override the definition
• enabled: false suppresses an inherited export

# loom defaults — archive for every thread
defaults:
  exports:
    - name: audit_archive
      type: parquet
      path: /archive/${thread.name}/${run.timestamp}/

# thread level — suppress for this thread
exports:
  - name: audit_archive
    enabled: false

Dynamic Path Naming

Export paths support context variables for unique-per-run outputs:

Variable	Description
${run.timestamp}	ISO 8601 UTC timestamp of execution start
${run.id}	UUID4 unique per execution
${thread.name}	Thread name
${weave.name}	Weave name
${loom.name}	Loom name

These are resolved at execution time, not config load time.

Error Handling

Each export has an on_failure setting (default: warn):

• warn — log warning, continue to next export, thread succeeds
• abort — raise ExportError, thread fails, remaining exports skipped

Exports only execute after a successful primary write, watermark persistence, and assertions. Export failures never corrupt the primary target or watermark state.

Run Context Variables

Two new system-wide context variables are available in audit column expressions and export path templates:

• ${run.timestamp} — ISO 8601 UTC timestamp
• ${run.id} — UUID4 identifier

These are generated once per execute_thread() call, ensuring consistent values across audit columns and export paths within a single execution.

Enhanced Flow SVG

The thread flow diagram now shows the complete data pipeline:

• Validation gate — appears after transforms when validations are configured
• Quarantine destination — branches down from validation
• Assertions checkpoint — appears after the target node
• Export destinations — fan out from the preparation stage as siblings of the target (not downstream of target)

When no validations, assertions, or exports are configured, the flow remains unchanged.

Sankey-Style Waterfall

The horizontal bar chart waterfall has been replaced with a proportional band-width Sankey flow:

• Rows read → after transforms → validation split → target + exports
• Quarantine branch shown with proportional band width
• Export bands shown alongside the target band
• Row counts labeled on each band

Observability

• explain() lists exports with name, type, and target path
• summary() shows per-export results (rows, status, errors)
• ThreadTelemetry.export_results captures ExportResult for each export (name, type, target, rows_written, duration_ms, status, error)

New Public Exports

From weevr.operations:

• write_export — write a DataFrame to an export target
• resolve_exports — resolve context variables in export paths
• resolve_export_path — substitute variables in a single path

From weevr.telemetry:

• ExportResult — per-export write result model

From weevr.model:

• Export — export domain model

From weevr.errors:

• ExportError — raised on on_failure: abort export failure

Compatibility

No breaking changes. All new fields default to None or empty — existing configurations work without modification.

Component	Version
Python	3.11
PySpark	3.5.x
Delta Lake	3.2.x
Microsoft Fabric Runtime	1.3