# Phase 21B.2 - Story Intelligence Experience Composition Refinement

## Purpose

Phase 21B.2 refines the development-only Story Intelligence Experience prototype so the Living Story View reads as one connected narrative map rather than separate character, location and asset regions.

This phase remains visual-only. It does not connect real Story Intelligence data, add SignalR, alter the import workflow, expose production navigation, create SQL, or modify the database.

## Composition Changes

- The central map container border and background were greatly softened.
- Characters, the active location and significant assets now share a single visual canvas.
- The scene title was reduced so the map has more vertical room.
- Category labels remain only as faint positional guides.
- The background now carries a subtler star/network texture behind the entities.

## Circular Location

The active location is now rendered as a circular node with the same broad visual language as character and asset nodes:

- Circular image mask
- Soft illuminated border
- Location name and secondary description below the node
- Stable `data-location-id` connection target

The existing prototype SVG location artwork is still used and masked with `object-fit: cover`, so future curated images can replace it without changing the layout contract.

## Circular Assets

Assets in the main Living Story View now render as circular image-first nodes:

- Circular image/avatar
- Label and state text underneath
- Size influenced by simulated narrative weight
- Stable `data-asset-id` connection targets

The compact discovery feed indicators were left unchanged because they serve a different scanning purpose in the left panel.

## Connection Lines

The SVG connection layer now draws multiple current-scene connection categories:

- Character relationship
- Character present at location
- Asset found at location
- Character knowledge or evidence connection to assets
- Asset-to-asset supporting evidence link

Connections use curved paths and category-specific styling. Supportive, uncertain, conflict, presence, evidence and knowledge lines are distinguishable by colour and line style rather than colour alone.

## Inline Relationship Labels

The central map now renders one or two inline relationship labels per scene on selected character-to-character connections. These labels summarize the active relationship state, while the right-hand Relationship Changes panel continues to show the latest detected changes.

Examples include:

- `trust strengthened`
- `tension increasing`
- `family`
- `questions`
- `motive exposed`

## Analysis Stage Progress

A compact analysis-stage rail was added to the header. It advances with simulated scene state and uses live-analysis stage labels:

- Reading
- Chapters
- Scenes
- Content
- Characters
- Locations
- Assets
- Relationships
- Knowledge
- Continuity
- Finalising

The real Story Intelligence pipeline classes were inspected, but not changed. This rail is prototype-only and maps the existing simulated scene `stage` and `progressPercent` values onto a compact display.

## Responsive Behaviour

The main desktop target remains 1920 x 1080 through 1366 x 768. At narrower widths the stage rail can scroll horizontally and the existing stacked layout is retained. Full mobile optimisation remains deferred.

## Limitations

- The map positions are still authored prototype positions, not graph-solved layout.
- SVG connection labels are intentionally sparse to avoid turning the experience into a dense generic graph.
- Location and asset artwork is still prototype SVG artwork, not final curated imagery.
- No historical replay/event store exists yet.

## Deferred

- Real data binding
- Durable Story Intelligence snapshot contract
- Production route/navigation
- Final image library
- Mobile-specific composition pass
- Phase 21C real-data integration