Skip to content

Marivo 0.4.0

Marivo 0.4.0 adds first-class event and lifecycle analysis, gives agents one public help coordinator, and strengthens the contracts that connect semantic authoring, execution, evidence, and repair.

  • Events and lifecycle state models are first-class semantic objects. Authors can define typed participants, event timestamps, lifecycle states, transitions, inception rules, and validity constraints, then load and inspect them through the semantic catalog.
  • Analysis now includes typed subjects and cohorts, event occurrence matching, sequences, funnels, time-to-event analysis, lifecycle replay, dwell and transition analysis, violation detection, funnel comparison, and funnel-loss attribution.
  • Event and lifecycle results carry explicit completeness, subject, time, and semantic-input evidence. Quality checks cover event identity, participant coverage, ordering, lifecycle transitions, censoring, and replay continuity.
  • marivo.help(...) is the single public help coordinator across datasource, semantic, and analysis. Focused topics, live object briefings, result contracts, and typed errors now lead agents to mechanically valid next steps.
  • Telemetry output rotates into bounded daily segments and prunes historical files by age and total size.
  • Use marivo.help() and qualified topics such as marivo.help("semantic.event") or marivo.help("analysis.events"). The previous md.help, ms.help, mv.help, and help_text entry points are no longer public.
  • Event and lifecycle operations require exact typed semantic references, explicit subject axes, and explicit completeness declarations where a bounded observation cannot prove completeness. Bare names and silent inference are rejected.
  • Public frame and result contracts use closed, kind-specific semantic input and evidence shapes. Code that depended on internal or generic payload fields must move to the documented frame, .show(), .contract(), and evidence APIs.
  • Semantic authoring help now owns constructor constraints and produces evidence-backed repair guidance for invalid or incomplete declarations.
  • Analysis planning preserves metric value columns, cross-entity grain requirements, subject axes, time scopes, and evidence continuity more consistently across operators.
  • Public frame representations, summaries, quality reports, and errors are more bounded and deterministic, with clearer continuation hints.
  • Cross-entity observations now explain required grain repairs instead of leaving agents to infer missing joins or grouping.
  • Help examples, live contracts, public exports, and packaged agent skills are checked together to reduce drift between the installed runtime and guidance.
  • Local telemetry no longer grows indefinitely in one events.jsonl file.
  • Added English and Chinese guidance for the unified help entry point, event and lifecycle semantic authoring, analysis workflows, evidence, and readiness.
  • Added design and implementation specifications for event journeys, lifecycle replay, ontology extensions, and the agent-facing public surface.
  • Created an isolated bilingual v0.4 documentation snapshot.
Terminal window
pip install marivo

Install the backend extra for each datasource you plan to use. Marivo 0.4.0 requires Python 3.12 or newer.