RTYI Documentary

Beyond the cartridge: A research and preparation workspace for the Return to Yoshi's Island documentary.

Quick Start · Analysis Pipeline · Project Structure · rtyi.land

About

This repository is the working hub for a long-form documentary about Return to Yoshi's Island, the Mario 64 ROM hack led by Kaze Emanuar. It holds everything needed to prepare for filming: interview question sets, narrative planning, stream evidence, contributor profiles, and curated quotes.

Planned interviews include:

• Kaze Emanuar
• Biobak
• Badub
• Kaze and Zeina together

Quick Start

Prerequisites

• Node.js 22+
• pnpm

Installation

git clone git@github.com:johannschopplich/rtyi-doc.git
cd rtyi-doc
pnpm install

Documentation Site

pnpm docs:dev       # Start local dev server
pnpm docs:build     # Build the site
pnpm docs:preview   # Preview the built site

Transcript Analysis

Raw stream transcripts live in transcripts/. Running the analysis pipeline extracts structured findings from each one and writes them to .data/streams/ as JSON. The synthesis step then aggregates those findings into documentary-ready content.

Create a .env file in the repository root with your API keys, then:

pnpm stream-analysis    # Per-stream extraction → .data/streams/
pnpm stream-synthesis   # Cross-stream aggregation → .data/synthesis/

Quality Checks

pnpm lint
pnpm format:check
pnpm test:types

Project Structure

rtyi-doc/
├── docs/                         # VitePress documentary research site
│   ├── .vitepress/               # Site config, theme, data loaders
│   ├── drafts/                   # Narrative arcs and chapter planning
│   ├── interviews/               # Per-person interview question sets
│   ├── synthesis/                # Documentary prep (generated from analysis)
│   ├── streams/                  # Stream pages and dashboard
│   ├── topics/                   # Findings grouped by documentary theme
│   ├── team/                     # Contributor profiles
│   ├── prompts/                  # Prompt and extraction documentation
│   ├── research/                 # Background research on documentary craft
│   ├── public/                   # Static assets
│   └── index.md
├── scripts/
│   ├── stream-analysis.ts        # Per-stream transcript extraction
│   └── stream-synthesis.ts       # Cross-stream aggregation
├── src/
│   ├── analysis/
│   │   ├── prompts.ts            # Prompt templates for stream extraction
│   │   ├── schemas.ts            # Zod schemas for analysis output
│   │   └── runner.ts             # Transcript processing and JSON output
│   ├── synthesis/
│   │   ├── prompts.ts            # Prompt templates for aggregation
│   │   ├── schemas.ts            # Zod schemas for synthesis output
│   │   └── runner.ts             # Aggregation execution
│   ├── stt-corrections.ts        # Speech-to-text cleanup rules
│   ├── constants.ts              # Paths and model defaults
│   └── utils.ts                  # Provider and model helpers
├── transcripts/                  # Raw stream transcript files (.txt)
├── .data/                        # Generated analysis artifacts
│   ├── streams/                  # Per-stream extraction output (JSON)
│   └── synthesis/                # Aggregated documentary prep (JSON)
├── package.json
├── pnpm-workspace.yaml
├── wrangler.toml
└── README.md

Analysis Pipeline

The pipeline has two stages. The first reads each raw transcript and extracts structured findings – development decisions, context and motivation, contributor roles, key stories, and open questions for follow-up interviews. Output goes to .data/streams/ as one JSON file per stream.

The second stage aggregates all per-stream output into documentary-ready material in .data/synthesis/:

• Story Arcs – arc-first narrative stories with embedded interview questions and quotes
• Narrative Arcs – thematic filming roadmap forming the documentary's high-level structure
• Topic Arcs – per-topic narrative summaries tracing how each theme evolved across streams

VitePress data loaders in docs/synthesis/ read these JSON files and render them as browsable pages on the research site.

Deployment

The site deploys to Cloudflare as a static build. Routing and asset configuration live in wrangler.toml, serving the site at rtyi.land.

License

This project is for internal documentary research and preparation. Content related to the RTYI team and project is used with permission for documentary purposes.