Changelog

Added source-to-clean-text TextSpan mappings to parse structure results

Added strict 0.9 standalone scene separators with x-strong break rendering and escaped literal support

Added strict diagnostics and remediation hints for directive fences shorter than three colons

Added support for application-defined audio URIs and strict playback attribute validation

Added generic, SSML 1.1, and provider targets with BCP-47 language preservation and explicit loss policies

Added Python 3.14 to the supported test matrix

BREAKING: Changed rate and pitch scales and added per-voice defaults with prosody transition diagnostics

BREAKING: Changed SSMD to strict 0.9 syntax with portable front matter, canonical formatting, migration, and trusted extensions

Changed CLI inference: plain .md needs ssmd_version or –from; .ssmd.md is the recommended complete-document name

Changed parsing, formatting, and migration to preserve paragraph semantics across adjacent fenced voice scopes

Documented API and CLI reference pages using Sphinx-compatible directives and tables

Documented strict 0.9 examples and clarified current authoring and legacy compatibility guidance

Added semantic and pronunciation language scopes with validated document routing hints

Added structural sentence-span integration for downstream TTS pipelines

Added sentence-neutral parsing with clean-text spans, structural events, and YAML front matter for TTS pipelines

Added compact V/R/P prosody attributes and symbolic prosody aliases for SSMD input

Improved sentence splitting diagnostics and model selection through phrasplit 0.3.5

Added compact voice blocks, recursive voice-binding defaults, and semantic round-trip validation

Fixed parsing of float numbers

Added release-tag version checks for wheels and source distributions

Fixed duplicated sentence-boundary pauses during SSML-to-SSMD round-trips

Preserved whitespace before leading decimals during parsing and sentence rendering

Added strict formatter, SSML, and CLI regression coverage for pauses between sentences

Added installed-wheel smoke validation outside the source checkout

Added command-line validation and SSMD/SSML conversion workflows with parser integration

Added parser, document, capability, diagnostics, and security coverage for SSMD processing

Added an external SSMD authoring skill with a validated command-line workflow

Added stable CLI JSON envelopes, command discovery, authoring configuration, voice workflows, and contract tests

Added portable YAML front matter with voice bindings, pause defaults, and provider-aware authoring

Added create JSON contract coverage for domain state, atomic output behavior, and header materialization

Changed title to portable string metadata excluded from plain text and spoken SSML output

Changed the SSMD agent workflow around result state, output inspection, and metadata warnings

Improved Windows compatibility for CLI and documentation workflow checks

Fixed CLI test and generated release-document behavior on Windows

Fixed sentence splitting so punctuation inside emphasis spans remains balanced across formatting and SSML conversion

Documented the authoritative project guidance and examples in Markdown and MyST sources

Added audited release metadata and reproducible validation gates for the 0.8.0 package