Documentación

Forward-looking research for attachment extraction and retrieval: provider boundaries, persistence, indexing, chunking, and phased delivery.

Two ways to get started with Byline: add it to an existing TanStack Start app with the CLI, or clone this repo and run the example application in dev mode.

The thinking behind Byline: why it exists, the problems it sets out to solve, and what it means to build a content platform for an era in which both people and machines read your content.

The load-bearing design decisions behind Byline: universal EAV storage, immutable versioning, the document-level vs version-level split (path, available locales, tree edge), patch-based updates, the schema/admin split, and the authorization model. Each links to its full reference.

Defining collections in Byline: defineCollection, the workflow system, lifecycle hooks (beforeCreate, beforeRead, afterUpdate…), and the admin presentation split.

How content leaves Byline and reaches its readers through the typed client SDK, server-function transports, markdown and llms.txt exports, future MCP support, and application caching.

How Byline provides ranked, multilingual site search through one provider contract, with built-in PostgreSQL and MySQL implementations that reuse the application's existing database.

How Byline controls who can do what, and how it records what was done: the actor and ability model, two-layer access control with read-side row scoping, and the version and document-level audit trails.

Byline separates interface translation from content translation, and treats both independently from the language a document is published in. An overview of the three i18n axes (host interface, admin interface, and content locales) and how they stay decoupled.

The building blocks of Byline's admin interface: the framework-agnostic UI Kit it is built from, and how the admin/editor configuration is registered into a host application without leaking into its public bundles.

Exact configuration, collection, field, and Client SDK contracts, with links to the subsystem references that own search, authentication, uploads, rich text, routing, and storage.

Two suites, two commands: pnpm test (unit, no database) and pnpm test:integration (real PostgreSQL and MySQL). Isolation, safety guards, and how to run a single test file.