API reference¶

Return a deep copy that shares nothing mutable with the original.

Return a copy with a new flux array (and optionally a new flux unit).

Return the sub-spectrum with wmin <= wavelength <= wmax.

Return a stable SHA-256 hex digest of the spectrum's numeric content.

Covers wavelength, flux, uncertainty (when set), mask (when set) and flux_unit. The digest is memoised per array identity: a second call on the same arrays is O(1); replacing any array (what every algorithm does) recomputes it. In-place mutation is not detected — see the class docstring.

Return a small JSON-safe summary, suitable for an MCP tool response.

Return the full spectrum as JSON-serialisable lists.

Append a processing step to the audit trail.

Return a deep copy (copy-on-write friendly): algorithms can mutate freely.

Return a SHA-256 digest of the context's scientific content.

The history is intentionally excluded so that a re-run with identical inputs and parameters produces the same hash. Covered: spectrum, spectra, image, images, metrics, line_fits, embedding, plus — only when non-empty, so contexts without them keep their historical digests — extras (JSON-serialisable values, kernel types carrying a content_hash, and ndarrays; anything else is skipped), the periodograms / light_curves arrays and the exports bytes. figures and catalog_lookups are not hashed.

True when no scientific payload has been set yet.

Return a compact, JSON-serialisable description of what the context holds.

Return a JSON-serialisable representation of the step.

Return a JSON-serialisable representation of the fit.

Return an :class:AlgorithmCategory from itself or its string value.

Perform the work. params is already merged with defaults and validated.

Read inputs from ctx, write outputs back into ctx, and return an :class:AlgorithmOutput summary. Subclasses must override this.

Return default_params overlaid with params; reject unknown keys.

Raise :class:InvalidParameterError if a required parameter is missing.

Raise :class:InputRequirementError if a required context input is absent.

Validate, run, time, and record one execution against ctx.

Always returns an :class:AlgorithmOutput — exceptions raised inside :meth:run are caught, converted to a failed result, and recorded in the context's history. Use :func:spectro_kernel.registry.run_algorithm if you prefer exceptions to propagate.

Return this algorithm's catalogue entry as a JSON-serialisable dict.

Build a successful result.

Build a failed result carrying error.

Return a JSON-serialisable representation.

Run every step against ctx (a fresh context is created when omitted).

Steps run in declared order unless any of them carries depends_on, in which case the pipeline is treated as a DAG and executed in topological order. With stop_on_error (the default) execution halts at the first failed step; otherwise it runs every step and reports overall success as the AND of all.

Return a JSON-serialisable description of the pipeline definition.

Set the pipeline name (and optionally description/version).

Append a step running algorithm with keyword params.

Pass depends_on=["other_label"] to declare a non-linear dependency — any step with depends_on switches the pipeline to topological-order execution.

Append a pre-built :class:PipelineStep.

Populate the builder from a preset-style config dict.

Expected shape::

{"name": ..., "description": ..., "version": ...,
 "steps": [{"algorithm": "snr_der", "params": {...}, "name": "..."}, ...]}

Populate the builder from a named YAML preset.

Return the assembled :class:Pipeline.

With validate (the default) every referenced algorithm must already exist in the registry, so typos fail fast at build time rather than mid-run.

Return a JSON-serialisable summary of the run.