WWDC Sessions — Agent-Native Knowledge Base

An agent-native index of Apple WWDC developer sessions (2014-2026, plus Tech Talks and Meet
with Apple): clean transcripts, structured metadata, inline code snippets, and links to the
documentation each session references. Built so AI agents (and humans) can consume WWDC content
without scraping JavaScript-rendered pages.

• 1638 sessions across 15 events · 1551 with full transcripts
• Machine entrypoint: catalog.json · also events.json · topics.json · llms.txt
• JSON Schemas: schema/ · Agent guide: AGENTS.md
• Per session: metadata.json, README.md, transcript.md, transcript.json

Quick start (agents)

GET catalog.json                                   # full index; has rawBase + a `path` per session
GET <rawBase>/<path>transcript.md                  # the talk, timecoded
GET <rawBase>/<path>transcript.json                # { id, language, source, wordCount, segments:[{start,text}] }
GET <rawBase>/<path>metadata.json                  # structured + resource links

Each linked Apple doc includes a sosumiURL (clean Markdown) so an agent can fetch documentation
on demand. See AGENTS.md for path/locator conventions.

Regenerate & validate

python3 scripts/build.py        # resolves Apple's rotating feeds dynamically; skips if unchanged (--force to override)
pip install -r requirements.txt
python3 scripts/validate.py     # catalog <-> filesystem integrity + JSON Schema validation

A scheduled GitHub Action refreshes the data and runs validation.

Provenance & license

Session transcripts, metadata, and code snippets are © Apple Inc., sourced from Apple's public
developer-video feeds for indexing and developer convenience. This project is free, non-commercial,
and complementary; every session links back to its source. WWDC and Apple are trademarks of Apple
Inc.; this is an independent index, not affiliated with or endorsed by Apple. The tooling in
scripts/ is MIT licensed (see LICENSE and NOTICE).