Marivo 0.4.16

Marivo 0.4.16 makes discovery easier to configure and recover, records the canonical value column used by each objective, and gives evidence page limits structured, copyable repairs. It also documents a progressive analysis help topology for agent-facing workflows.

Highlights

Focused marivo.help("analysis.<target>") now exposes the strategy contract for every discover.* objective: its default, legal alternatives, applicable semantic kinds, threshold meaning, and when-to-use guidance. The strategy facts are rendered from the runtime registry.
discover now resolves an omitted value according to the frame: a DeltaFrame uses its canonical delta column, while a MetricFrame uses its declared measure/value column. The resolved column is recorded in params and evidence, so omitted and explicit values are equivalent.
session.evidence.findings() and .digests() now document bounded limits in [1, 100] (defaults 50 and 10), immutable page state, and has_more/next_cursor continuation. Invalid limits raise EvidenceLimitError with a copyable repair.

Changed

Explicit discover value selection validates existence and numeric type, and failures include candidate numeric columns and a retry path. A multi-metric frame should be projected with frame.metric(...) before discovery.
Discovery strategy guidance distinguishes single-point anomalies, local regime changes, whole-window behavior, and sustained shifts that belong to compare.

Fixed

An observe → compare → discover chain no longer treats the paired baseline bucket as a dimension when resolving current-frame axes.
Omitted value no longer fails on a DeltaFrame merely because its current, baseline, delta, and percent-change columns are all numeric; a MetricFrame with an additional numeric axis can likewise use its declared measure.

Documentation

Updated the analysis workflow and evidence documentation in English and Chinese with discovery strategies, value resolution, and bounded pagination.
Documented the progressive analysis live-help topology and its ownership boundaries in the analysis design specification.

Internal

Added regression coverage for real observe/compare/discover chains, focused strategy help, evidence paging, and structured limit repairs.

Installation

curl -fsSL https://marivo.io/install.sh | bash

In mainland China, use:

curl -fsSL https://marivo.io/install-marivo-cn.sh | bash

For manual installation:

pip install marivo

Marivo 0.4.16 requires Python 3.10 or newer.