@aceshooting/lyra-ui 6.2.0 → 7.1.0

package/CHANGELOG.md

@@ -1,5 +1,86 @@
 # Changelog
 
+## 7.1.0
+
+### Minor Changes
+
+- bf9ac61: Resolve public integration gaps:
+
+  - add the side-effect-free `@aceshooting/lyra-ui/localization.js` runtime entry;
+  - add `--lr-icon-button-border-hover`, falling back to the base border;
+  - document and test the reflected `ariaControlsElements` contract for menu triggers;
+  - make popover/dropdown `aria-controls` target their public host so native and Lyra triggers can
+    resolve it;
+  - make tooltip and checkbox descriptions resolvable across their trigger/control shadow
+    boundaries; and
+  - accept MapLibre GL JS v5 or v6, with version-specific worker guidance.
+
+## 7.0.0
+
+### Major Changes
+
+- fb14fa1: Make the approved breaking contract corrections for the next major release.
+
+  - `<lr-checkbox-group>` now consumes each child's native-style events and emits exactly one
+    group-owned `input`, `change`, and `lr-change` sequence with `{ value: string[] }`. It also scopes
+    selection to checkboxes owned by that group and silently resynchronizes its value and form data
+    after programmatic child changes. Consumers that listened for leaked child events on the group
+    must instead listen on the child checkbox itself.
+  - `<lr-diff-view>` now defaults `maxLines` to `5000` and renders a localized fallback above that
+    ceiling. Set `maxLines` to `Infinity` to preserve uncapped behavior; the line diff now uses a
+    linear-space algorithm, but exceptionally large diffs can still be expensive.
+  - `<lr-popover>` now returns focus to its trigger consistently after light dismiss and
+    programmatic close, matching its Escape behavior. Use `hide({ focusTrigger: false })` when a
+    programmatic close must leave focus elsewhere.
+  - Remove the never-emitted `lr-highlight-activate` event declarations from `<lr-code-block>`,
+    `<lr-code-block-core>`, and `<lr-notebook-viewer>`. Listen to the viewer's documented anchor and
+    text-selection events instead.
+  - `<lr-stepper>` now exposes list/progress-navigation semantics with
+    `aria-current="step"` instead of incomplete tab semantics without associated tab panels.
+    Selected-step state and keyboard activation remain available.
+  - Toggleable `<lr-chip>` instances now put their toggle semantics on a separate native control;
+    default-slot content is an inert label rather than an unrestricted interactive subtree. Move
+    links and buttons outside a toggleable chip.
+
+### Minor Changes
+
+- fb14fa1: Expand the public component contracts needed by advanced consumers.
+
+  - Export the document-viewer anchor, highlight, search, selection, and target types through the
+    owning granular component entry, with type tests covering those imports.
+  - Complete native-style `input`/`change`, focus, selection, and editing contracts across the
+    affected form and conversation controls, including emoji picker and token input.
+  - Add the documented viewer navigation, search, highlight, comparison, preview, and theme hooks,
+    plus the corresponding component parts and custom properties.
+  - Complete the typed agent-evaluation, evaluation-dataset, retrieval, and data-view surfaces that
+    previously required consumers to infer internal shapes.
+
+- 19022c7: Resolve the current feature-request backlog:
+
+  - add configurable no-flash theme bootstraps and a Lit-free gemstone palette entry;
+  - expose a pre-mount chart `seriesPalette()` helper and document its theme-token indirection;
+  - let `lr-app-rail` select persisted fields, including `preferredMode`, without restoring transient
+    mobile-open state;
+  - add `lr-icon-button` border and hover-foreground tokens;
+  - size gemstone swatches from their fill token and keep `lr-table`'s unnamed-grid warning out of
+    production; and
+  - document the supported SheetJS CDN install path and the unsafe npm-audit downgrade suggestion.
+
+### Patch Changes
+
+- fb14fa1: Correct accessibility, localization, lifecycle, security, responsive-layout, and rendering defects
+  across the component families.
+
+  - Reconcile accessible names, stateful ARIA, roving focus, focus return, wrapped-child event
+    suppression, live regions, disabled behavior, and native control forwarding.
+  - Make inherited locale and direction reactive through composed trees, use locale-aware text
+    folding and sparse highlight offsets, and localize the remaining viewer and status messages.
+  - Harden remote viewer loading, sanitization, generation ownership, size/resource guards, and
+    reconnect behavior while preserving empty and error states.
+  - Fix container-responsive layouts, hover/focus parity, reduced-motion behavior, theme-token
+    resolution, and viewer allocation/geometry updates.
+  - Avoid quadratic DOM walks when painting capped DOCX search matches.
+
 ## 6.2.0
 
 ### Minor Changes

package/README.md

@@ -27,7 +27,7 @@
 **Lyra UI — the free, independent web-component alternative.** A MIT-licensed [Lit](https://lit.dev)
 library for accessible forms, dashboards, charts, data visualization, and Conversation & Agent UI.
 It is a practical open-source alternative to [Shoelace](https://shoelace.style/) and
-[Web Awesome](https://webawesome.com/), with 256 custom elements, native custom-element APIs,
+[Web Awesome](https://webawesome.com/), with 268 custom elements, native custom-element APIs,
 tree-shakeable imports, its own `--lr-*` design tokens, built-in localization and RTL support,
 and no runtime dependency on either project.
 
@@ -55,6 +55,9 @@ npm install @aceshooting/lyra-ui
 # optional peer: maplibre-gl, only needed for <lr-map> — also import
 #   `maplibre-gl/dist/maplibre-gl.css` yourself once, since lr-map only
 #   ships its own legend/popup chrome CSS, not maplibre-gl's own stylesheet.
+#   Lyra supports MapLibre v5 and v6. v5's standard build includes its worker;
+#   v6 is ESM-only, requires WebGL2, and also needs its module-worker URL
+#   configured for your bundler before the first map is constructed.
 #   <lr-map> falls back to OpenStreetMap's demo tile server when you don't
 #   set `mapStyle` — fine for local dev, but NOT for production (no capacity
 #   guarantees, requires an identifying User-Agent, subject to IP-blocking —
@@ -62,6 +65,20 @@ npm install @aceshooting/lyra-ui
 #   apps must supply their own `mapStyle`.
 ```
 
+For example, Vite consumers using MapLibre v6 configure its self-contained worker once:
+
+```js
+import { setWorkerUrl } from 'maplibre-gl';
+import workerUrl from 'maplibre-gl/dist/maplibre-gl-worker.mjs?worker&url';
+import 'maplibre-gl/dist/maplibre-gl.css';
+
+setWorkerUrl(workerUrl);
+```
+
+Webpack, esbuild, Rollup, and direct-browser ESM use different worker URL forms; follow
+[MapLibre's ESM installation guide](https://maplibre.org/maplibre-gl-js/docs/#esm) for the matching
+setup.
+
 ## Usage
 
 Import just what you use (tree-shakeable, granular entry points):
@@ -249,13 +266,20 @@ Markdown, phone-number metadata, and flag artwork remain optional peer dependenc
 `aria-label`s) resolves through a small runtime, in two ways you can combine:
 
 ```ts
-import { registerLyraLocale, setLyraLocale } from '@aceshooting/lyra-ui';
+import {
+  registerLyraLocale,
+  setLyraLocale,
+} from '@aceshooting/lyra-ui/localization.js';
 
 // App-wide: register translations once, anywhere in the app.
 registerLyraLocale('fr', { close: 'Fermer', retry: 'Réessayer', /* ... */ });
 setLyraLocale('fr'); // or just set <html lang="fr">/an ancestor `lang` — components pick it up
 ```
 
+The dedicated `localization.js` entry is side-effect-free: it does not register the component
+graph. The package root continues to re-export the same runtime for compatibility, but importing
+it also registers the non-peer-gated components.
+
 ```html
 <!-- Per-instance: override specific keys on one element without a global registry. -->
 <lr-toast .strings=${{ close: 'Fermer' }}></lr-toast>
@@ -354,7 +378,7 @@ coverage automatically from the bundled `web-types.json` — JetBrains IDEs pick
 
 ## Components
 
-The catalog below lists all 256 tags in the current Custom Elements Manifest, grouped by
+The catalog below lists all 268 tags in the current Custom Elements Manifest, grouped by
 capability. The manifest and live docs are the authoritative sources for the complete generated
 API details.
 
@@ -471,8 +495,11 @@ API details.
 | `<lr-retrieval-results>` | — (extra) | Orchestration-level ranked-chunk-list surface — deduplication, optional grouping by source, multi-selection, pagination/infinite loading, and a compact/expanded presentation switch; composes an `lr-chunk-inspector` per row and an internal `lr-virtual-list` for large result sets |
 | `<lr-retrieval-trace>` | — (extra) | Retrieval pipeline's stage timeline (query rewriting, embedding, retrieval, reranking, filtering) rendered through `lr-span-waterfall`, plus a disclosure list exposing each stage's evidence (chunks via `lr-chunk-inspector`, free-form text, and/or metadata); never fetches or computes retrieval results itself |
 | `<lr-grounding-summary>` | — (extra) | Claim-level scorecard for one generated answer — supported/unsupported claim counts, citation coverage, an optional confidence score, warnings, and (when supplied) a list of evidence citations; composes `lr-stat` and `lr-citation-badge`, pure projection and event conduit |
+| `<lr-claim-evidence>` | — (extra) | One claim's support verdict, score, citations, and evidence excerpts, with typed citation activation for opening the original source |
 | `<lr-context-inspector>` | — (extra) | Inspection view of the exact context assembled for a model call — per-segment token estimates via `lr-context-meter`, source attribution via `lr-citation-badge`, copy/export affordances, and truncation-boundary/redaction-marker rendering; pure projection, never fetches, estimates, or redacts itself |
 | `<lr-rag-answer>` | — (extra) | Grounded answer surface composing sanitized Markdown, claim-level grounding summary, citation badges, and source cards; controlled, localized, and emits citation/retry events without fetching data |
+| `<lr-retrieval-compare>` | — (extra) | Side-by-side comparison of retrieval runs, including latency, hit count, score, and overlapping source identifiers; controlled and computation-free |
+| `<lr-rag-eval-dashboard>` | — (extra) | Aggregate RAG quality scorecard for groundedness, citation coverage, relevance, recall, latency, and cost, with per-case drill-in events |
 
 **Knowledge base & document management**
 
@@ -490,7 +517,11 @@ API details.
 |-----------|---------|-------|
 | `<lr-agent-run>` | — (extra) | Top-level shell for one `AgentRun` — lifecycle-status badge, elapsed time, current step, model/cost summary, and built-in Cancel/Retry controls in a header, plus `tasks`/`tools`/`reasoning`/`output` composition slots; a shell only, every piece of per-step rendering routes through an existing primitive |
 | `<lr-agent-workspace>` | — (extra) | Responsive controlled AI workspace composing the transcript, composer, agent run, tool timeline, retrieval results, grounding summary, and context inspector, with replaceable message/details/composer slots |
+| `<lr-subagent-panel>` | — (extra) | Parent/child agent hierarchy with lifecycle status, task summaries, and typed selection/cancel events; the host owns orchestration |
 | `<lr-agent-trace>` | — (extra) | Provider-neutral agent/LLM trace view — a span-kind filter row, a handoff quick-jump list, and the full trace hierarchy, all rendered through `lr-trace-tree` over one shared `LyraSpan[]` array |
+| `<lr-mcp-app>` | — (extra) | Sandboxed MCP Apps host with explicit resource URL, origin allowlist, CSP, and permission policy inputs; emits typed load, error, and message events |
+| `<lr-prompt-studio>` | — (extra) | Versioned prompt editing and testing workspace with model/variable inputs, controlled run events, and version-selection events |
+| `<lr-schema-viewer>` | — (extra) | Read-only, recursively expandable JSON Schema browser for tool inputs, structured outputs, and protocol payloads |
 | `<lr-tool-timeline>` | — (extra) | Chronological list of an agent run's tool/function calls, each rendered through `lr-tool-call-chip` (name/status/duration) and `lr-tool-result-view` (args/result), with per-entry retry counts, sensitive-field redaction, and a shared `lr-tool-approval-dialog` for entries gated behind human approval |
 | `<lr-memory-panel>` | — (extra) | Agent working-memory surface — short-term context and long-term memories, each item's confidence and optional grounding provenance via `lr-provenance-panel`, with add/remove/forget actions gated behind an `lr-confirm-bar` confirmation step |
 | `<lr-policy-summary>` | — (extra) | Read-only list of guardrail, permission, privacy, and tool-policy decisions (`allow`/`deny`/`needs-review`), each with an always-visible, accessible explanation never conveyed by color alone; composes `lr-badge` and `lr-callout`, with `lr-details` for optional richer detail |
@@ -547,7 +578,12 @@ each one-liner below.
 | Component | Mirrors | Notes |
 |-----------|---------|-------|
 | `<lr-chat-message>` | — (extra) | Role-based (`user`/`assistant`/`system`) message bubble shell; avatar/badges header, status-aware footer (built-in retry on `status="failed"`), attachments strip — renders no message content itself, just the chrome around a slotted body |
+| `<lr-message-parts>` | — (extra) | Provider-neutral renderer for ordered text, reasoning, tool-call, tool-result, citation, file, and declarative widget parts |
 | `<lr-chat-viewport>` | — (extra) | Transcript scroll container: stick-to-bottom while an answer streams, a "jump to latest" pill, and an unread divider; auto-detects slotted `<lr-chat-message>` children vs. a single nested `<lr-virtual-list>` (virtual mode) and defers scrolling to it |
+| `<lr-prompt-input>` | — (extra) | Form-associated multimodal prompt editor for text, attachments, source selection, and send/stop behavior, emitting a typed submission payload |
+| `<lr-prompt-queue>` | — (extra) | Controlled queue of pending prompts with reorder, edit, remove, and submit-next request events |
+| `<lr-selection-toolbar>` | — (extra) | Contextual toolbar for actions on selected transcript text, positioned from a host-supplied selection rectangle |
+| `<lr-realtime-session>` | — (extra) | Voice/realtime session controller surface that visualizes connection and speaking state and emits connect, disconnect, mute, and interrupt requests |
 | `<lr-message-actions>` | — (extra) | Per-message action toolbar for `<lr-chat-message>`'s `actions` slot — opt-in copy/regenerate/edit/feedback built-ins plus a default slot for custom controls (e.g. a slotted `<lr-branch-picker>`); `role="toolbar"` with APG roving-tabindex across every stop |
 | `<lr-message-feedback>` | — (extra) | Thumbs up/down for one assistant message with an optional inline reason-chips + free-text comment disclosure; emits, never persists — the host reflects a prior rating back via `value` |
 | `<lr-branch-picker>` | — (extra) | The "‹ 2 / 5 ›" navigator across regenerated/edited message variants; pure controlled (`lr-branch-change`, never mutates its own `index`) — same contract `<lr-pagination>` establishes for `page`; renders nothing while `count < 2` |