Marivo 0.5.1

Marivo 0.5.1 introduces a Run-first public Session runtime read surface and hardens the persistence and recovery paths behind it.

Highlights

session.runs() and session.get_run() expose a closed immutable lifecycle: IncompleteRun, SucceededRun, or FailedRun.
session.artifact(ref) restores one exact committed Artifact, while session.graph() projects bounded factual producer, consumer, reuse, head, failed, and incomplete state.
Every Artifact exposes finding_count, findings(), and finding(id). Finding audit is Artifact-owned and retains complete derivation data.
mv.session.inspect() now returns only a Session summary and bounded Run page. Runtime and Graph reads do not check current semantic authority or datasource freshness; use session.revalidate(ref) for the authority and Evidence-integrity check.

Canonical Session runtime truth is persisted before it is projected into Run, Artifact, Finding, and Graph reads. Nullable metric presence markers preserve the distinction between an absent value and a present null value.
Runtime reads are an atomic cutover. Frame/job reads, the Session Evidence namespace, digest pages, derivation traces, Finding-selection compatibility, and their Help routes have no aliases or migration path.
The packaged marivo-analysis skill and analysis Help routing now follow the same narrow Run-first ownership model.

Interrupted and incomplete runs recover from persisted state without fabricating terminal success or losing the bounded failure context.
Session read helpers now retain the correct runtime and session identity across recovery, replay, and graph projection.

Updated the English and Chinese Session runtime, evidence, analysis workflow, API, and packaged-skill guidance for the new read contracts.
Added the design documentation for canonical runtime truth and bounded Session graph reads.

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 0.5.1 requires Python 3.10 or newer.