Marivo 0.4.9

Marivo 0.4.9 makes analysis readouts more decision-first for agents, adds multi-metric quality assessment, and tightens the contracts for catalog and session continuations.

Highlights

Analysis artifact .show() and .render() output now leads with the current decision context, issues, evidence, and interpretation before the data preview. The default card is bounded to 50 rows or 8 KiB and reports exact omission counts with a copyable recovery call when more data exists.
session.assess_quality(...) accepts multi-metric MetricFrame artifacts and returns one joint report. Frame-level checks run once, measure-level checks run once per authoritative metric binding, and typed issues and findings retain the affected metric identity.
The analysis workflow now provides a minimal first-observation path and append-only step scripts. Later decision rounds restore exact input artifacts by ref instead of re-executing earlier scripts or relying on an implicit latest result.
Semantic catalog browsing exposes copyable receiver-qualified calls such as catalog.metrics.show() and catalog.metrics.get(...), while complete mechanical continuations remain available through .contract().show().

Changed

Artifact cards show only state-dependent continuation hints. Complete capability and semantic reacquisition details are kept in the artifact contract, where stable capability identifiers remain available in structured output.
Multi-metric quality checks validate the ordered, unique, materialized measure bindings before execution and include the metric identity in repair guidance and evidence scope.
Session-owned catalog entries are reacquired from the current session catalog before analysis operations, keeping immutable catalog snapshots from being mixed across sessions.

Fixed

Analysis continuation labels now render public receiver calls, including session.compare(...) and session.attribute(...), instead of exposing internal capability identifiers as the primary next step.
Bounded artifact cards no longer repeat full continuation matrices or semantic acquisition details, reducing duplicate and potentially misleading guidance.

Documentation

The analysis workflow and semantic-layer guides now document the decision-first card order, exact truncation recovery, multi-metric quality reports, catalog ownership, and append-only script discipline.

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

Install the backend extra for each datasource you plan to use. Marivo 0.4.9 requires Python 3.10 or newer.