# Lixity v0.7.3 Released 21 October 2026. This release connects manuscript reading, visible analysis and saved research more directly. It also consolidates research history, dossier navigation and interface typography, and strengthens archive exports and service inspection. ## Read the measurements in the manuscript Enable **Analysis** in the manuscript reader. The first activation selects **Sentence length** and marks counted text immediately. Open the named metric beside the switch to choose another measurement or turn its highlights off; your choices stay in the browser. Counted sentences, words, patterns or quoted speech can be marked directly in the prose. Hover a marked passage, tap it, or use the paragraph menu to inspect its literal evidence and paragraph measurement. Turn Analysis off to return to plain reading. Text marks and paragraph comparisons answer different questions: marks identify what was counted; paragraph colours indicate a difference from that chapter's reference. A lexical match does not establish a grammatical construction, narrative distance, literary quality or a need to rewrite. Missing comparisons, unmappable source projections and unsupported browser features stay explicit. Manuscript bytes and native text selection are preserved. The CLI exposes the same evidence from the shared counters: ```sh lixity profile manuscript.md --layer asl --evidence lixity profile manuscript.md --layer filter --evidence --json ``` The opt-in JSON contracts are `paragraph-layer/1` and `paragraph-layer-evidence/1`; default profile JSON v2 is unchanged. Evidence uses Unicode codepoint offsets and a SHA-256 identity for the actual paragraph text. Human output is bounded; JSON does not add another copy of the manuscript. ## Follow saved research without losing your place At a paragraph, **… → Saved research** opens the context beside the text. Follow saved claim and dossier versions, inspect the retained source and return to the passage. Explicit pins remain distinct from newer records. Unresolved quotations retain their review state; Lixity never silently accepts a changed quotation. Wide dossier views keep a heading tree to the right of the document. Narrow screens and companion views use a compact, reversible heading disclosure. Decision review shows the exact decision and dossier versions together, with existing explicit acknowledgement and concurrency guards. Research history presents the selected version and full change reason once, translates stored choices into readable labels and omits absent optional fields. Version navigation, technical provenance, meaningful zero/false values and pinned references remain available. No archive migration is needed. ## A quieter, more consistent workbench The workbench groups **Manuscript & Analysis**, **Research & Dossiers** and **Project & exports**; its tabs remain **Manuscript**, **Research** and **Project**. Navigation precedes project actions; repeated project headings and the generic analysis slogan no longer consume reading space. Shared typography distinguishes interface metadata, controls and content while keeping manuscript prose and code suited to their different purposes. Reader jumps account for nested scrollports and sticky controls. Dossier returns preserve their saved position and focus, including a previous manuscript return. Stale navigation and interrupted effects are cancelled. Touch-restored focus no longer opens an unrelated tooltip; keyboard and explicitly requested help remain available. Reduced-motion preferences are respected. The analysis overview consolidates secondary measurements and scene details into reversible groups. Long lists and wide tables remain keyboard-scrollable. Pacing labels name their ASL units and explain the existing Hook heuristic; zero and unavailable results stay distinct. Declared register/POV filters restrict the style table display and retain the whole-manuscript reference. They do not recalculate a within-voice baseline. ## Shorter commands, safer archive operations Analysis and export commands can reuse `manuscript = "drafts/manuscript.md"` in the selected root's `lixity.toml`. Use `--project` to select a project explicitly; explicit input paths take precedence. Research manuscript commands share the configured-path resolver while retaining their existing write and retention guards. See the [CLI reference](https://mfahsold.github.io/lixity/USAGE.md#default-manuscript) for command scope. Research exports copy one accepted snapshot, verify content addresses and preserve an existing backup on pre-replacement failure. Invalid destinations, oversized restore candidates, malformed envelopes and duplicate archive members fail explicitly. Failures after replacement report uncertain durability. SQLite index errors retain their actual cause and path. The immutable archive remains authoritative; its search index is rebuildable. `lixity service status` inspects existing service state without creating or repairing runtime directories or acquiring the lifecycle writer lock. ## Install or update For a new installation, install [uv](https://docs.astral.sh/uv/), then run: ```sh uv tool install --python 3.12 --managed-python lixity uv tool update-shell ``` Open a new terminal, then start the local app: ```sh lixity --version lixity service start --open ``` For an existing PyPI tool installation: ```sh lixity service stop uv tool upgrade lixity lixity --version lixity service restart --open ``` A pinned wheel installation needs an explicit replacement: ```sh lixity service stop uv tool install --python 3.12 --managed-python --reinstall-package lixity "https://github.com/mfahsold/lixity/releases/download/v0.7.3/lixity-0.7.3-py3-none-any.whl" lixity --version lixity service restart --open ``` Verify a downloaded wheel against the release's `SHA256SUMS`. Restart every managed instance with its port and reload browser tabs. Installation alone does not replace a running process; restart reuses recorded startup settings, which may differ from a project selected later in the browser. See [installation and recovery](https://mfahsold.github.io/lixity/INSTALLATION.md). ## Compatibility and limits Product versions, JSON schemas and archive versions are independent. Research is now classified as a supported, explicit-project capability rather than labelled experimental as a whole. This does not change retention consent, immutable revisions or the accuracy of authored claims. Individual integrations, dossier apply/export workflows and NDA drafts retain their documented limitations. Back up archives and keep their readers and writers compatible. English remains the default analysis language; automatic detection is explicit. Seven interface languages share the same controls. Resource and validation coverage still varies by analysis language. Built-in analysis stays local and does not rewrite prose. README, guides, screenshots and the [read-only example](https://mfahsold.github.io/lixity/demo/report.html) follow the same author workflow. Public Pages measurement remains separate from the local app, ignores query strings, fragments, forms and selections, and honours browser privacy signals. Missing statistics access means unknown counts, not zero visits.