RFC 013 Output Projection Profiles & Scoped Context

RFC 013: Projection Profiles

Status: Implemented

Summary

Vyasa currently supports a single output target: HTML. This proposal introduces “Projection Profiles” to allow:

1. Defining multiple output targets in vyasac.toml.
2. Organizing templates and context per target.
3. Keeping the global context.vy clean from target-specific configuration.

Motivation

Currently, all configuration lives in a single context.vy. This leads to pollution when you have commands specific to HTML (e.g., span { class="..." }) vs Voice or JSON. We need a way to scope context and templates to a specific output projection.

Proposal

1. vyasac.toml Configuration

Define projection targets in the project manifest:

[projections]
html = { path = "view/html" }
voice = { path = "view/voice" }
json = { path = "data/json" }

2. File Organization

Instead of a flat templates/ folder, use scoped directories:

workspace/
├── vyasac.toml
├── content/ ...
└── view/
    ├── html/
    │   ├── templates/ ... (HTML templates)
    │   └── context.vy (HTML-only commands)
    └── voice/
        ├── templates/ ... (Voice templates)
        └── context.vy (Voice-only commands)

3. Build Command syntax

Allow users to select a projection at build time:

vyasac build --projection html  # Uses view/html/context.vy + templates
vyasac build --projection voice # Uses view/voice/context.vy + templates

4. Scoped Context Loading

When building for html:

1. Load Global context.vy.
2. Load view/html/context.vy (This overrides or adds to global).
3. Templates in view/html/templates override default templates.

5. Update (2026-07): Stream Vocabulary Localization

To support localized, domain-accurate terminology across interlinear streams, each stream’s scoped context.vy can declare a vocabulary block. This dictionary maps structural container labels, entity display names (graph_nodes), and relational action/role labels (graph_edges) for viewer UI rendering:

content/mula-deva/context.vy

`vocabulary {
    structure = [ "skandha" = "स्कन्ध", "adhyaya" = "अध्याय", "shloka" = "श्लोक" ]
    leaf_default = "shloka"

    entities = {
        "arjuna" = "अर्जुन"
        "dhritarashtra" = "धृतराष्ट्र"
    }

    actions = {
        "uvacha" = "उवाच"
        "speaker" = "वक्ता"
        "note" = "टिप्पणी"
        "variant" = "पाठान्तर"
    }
}

During compilation (vyasac pack), these vocabulary maps are serialized into the publication manifest (under stream_vocabularies) and graph_dict, allowing breadcrumb navigation and graph badges to display localized labels without hardcoding.