Bahai Story Library

Editorial Process

The Bahai Story Library uses a source-trust model. Rather than manually editing every individual story, we approve sources upfront. Every story drawn from an approved source then flows through an automated publication pipeline that enforces our content policies.

1. Source approval (manual, one-time per source)

A source — a book, magazine archive, or other published work — is approved once and assigned a verification tier:

2. Automated ingestion

Once a source is approved, the ingestion pipeline parses it, extracts discrete stories, auto-tags each one (figures, themes, era, location, age band, etc.), generates a citation, and proposes age-appropriate discussion questions and reflection prompts.

3. Automated policy gate

Every story passes through a content-policy scan before publication. Failures route to a review queue. The scan checks:

4. Audio generation

Once published, the story enters the audio queue. The pipeline looks up the source author in our authors database and selects a matching voice (female author → female voice; male author → male voice; voice character chosen per author's era and personality). OpenAI TTS-HD generates the MP3, OpenAI Whisper aligns word-level timestamps for karaoke and captions, and the file is cached forever on Cloudflare R2.

5. Versioning and rollback

All content lives in Git. Every published story has a complete edit history. If a source is ever revoked or a story needs to be removed, it can be done in a single commit and the entire publication trail remains auditable.

What we will not publish