Built-in Operator Workflows

Audience: Python builders
Status: Preview

Built-in operators should feel like typed workflow steps, not generic command transport.

Design rule

Prefer this:

• object-first discovery
• typed operator configuration
• typed result objects
• explicit conversion to downstream DTOs only at the edge

Avoid making the primary user story about raw payload assembly.

Current log-to-AVO example

The current well/log example is log-to-AVO. It demonstrates the typed workflow style; it is not the whole operator identity.

1. discover a Wellbore
2. resolve elastic inputs through elastic_log_set(...)
3. define interval layering
4. define the AVO experiment
5. run AVO
6. export a response or crossplot source only when needed

from ophiolite_sdk.avo import (
    AngleSampling,
    AvoExperiment,
    ElasticChannelBindings,
    LayeringSpec,
)

elastic = wellbore.elastic_log_set(
    bindings=ElasticChannelBindings(vp="Dt", vs="Dts", density="Rho")
)

layering = LayeringSpec.fixed_interval(20, unit="ft")
experiment = AvoExperiment.zoeppritz(
    angles=AngleSampling.range(0, 40, 5)
)

result = elastic.run_avo(
    layering=layering,
    experiment=experiment,
)

response = result.response_source()
crossplot = result.crossplot_source()

The main point is that users talk about wells, elastic channels, layering, and experiments, not request envelopes.

Current seismic processing example

The current seismic example demonstrates a typed built-in operator chain over a project-owned survey asset:

1. open a project
2. resolve a survey asset
3. pick a section
4. build a typed processing pipeline
5. preview
6. run the pipeline and persist the output as a derived project asset

from ophiolite_sdk import Project, SectionSelection, TraceLocalPipeline

project = Project.open("demo-project")
survey = project.surveys()[0]
selection = SectionSelection.inline(120)

pipeline = (
    TraceLocalPipeline.named("Bandpass + RMS AGC")
    .bandpass(8.0, 12.0, 45.0, 60.0)
    .agc_rms(40.0)
)

preview = survey.preview_processing(selection, pipeline)
processed = survey.run_processing(
    pipeline,
    output_collection_name="derived-seismic",
)

Again, the public language stays close to what the user is actually doing.

TraceBoost Workbench library templates

The TraceBoost Workbench/reference application also ships a small built-in library of local processing templates for interactive exploration.

The current starter set is intentionally short:

• AGC RMS 250 ms
• Bandpass 10-20-60-80 Hz
• Envelope
• Similarity Tight
• Similarity Balanced
• Dip Balanced

These are app-level starter presets, not a separate SDK contract family. They are exposed through the processing workspace “Apply library template” control and then behave like ordinary saved local templates.

Loose-store compatibility lane

TraceBoostApp and SeismicDataset still exist for direct .tbvol and .tbgath workflows outside an OphioliteProject.

Use that lane when:

• the seismic store is not project-owned yet
• the workflow is import-first or desktop-local
• you intentionally want direct store-path control

Do not teach that lane as the primary builder story for project-owned seismic operators anymore.

External operators are a different surface

Do not mix built-in workflow composition with external operator authoring.

Use:

• typed workflow helpers for built-in runtime capabilities
• ophiolite_sdk.operators for external operator packages

That separation keeps the main SDK legible while still exposing extensibility as a first-class system.

Current non-goal

Charts, widget hosts, and app wrappers are not part of the main built-in operator story for now.

They consume workflow outputs. They do not define the public workflow vocabulary.

Next: Python SDK reference