Changelog

All notable changes to FeynmanLM are documented here. Dates are in YYYY-MM-DD format.

FeynmanLM follows semantic versioning against its public surface: the MCP tool contracts, the knowledge-base/sync format, and licensing compatibility. While the version is 0.x, minor releases may include breaking changes to that surface; patch releases never do.

0.1.0 — 2026-07-13

Versioning restarted at 0.1.0 to adopt proper semantic versioning. Earlier releases used a retired numbering scheme and have been unpublished; this release supersedes all of them (existing installs update to it normally).

Highlights:

• Added Meta's Muse Spark 1.1 as a text AI provider via the new Meta Model API — add a Meta API key in Settings and pick Meta for chat or background tasks (Meta's consumer chat app doesn't support MCP connectors, so it can't connect to FeynmanLM directly)
• Added Gemini as a Read Aloud provider, with overnight batch audio generation at half price
• Added one-click approval for MCP connections in the app and a redesigned pairing consent page
• Made Sign in with Apple the primary unlock: your license follows your Apple ID, with self-serve device management and key recovery
• Overhauled read-aloud reliability: per-paragraph generation with retries, synced playback position, sleep timer, AirPods/system media controls, and a playback scrubber
• Moved the Google Drive connector to file-picker scope and YouTube to a FeynmanLM-owned OAuth client
• Hardened the MCP server: hosted relay only, stricter auth, and friendlier client names
• Added an AI support agent on the website that can file GitHub issues for licensed users
• Added custom source ordering in Project pools and read-status/content-status filters in the sources sidebar
• Fixed several main-thread hangs (app launch, sidebar, EPUB chapter loads, thumbnail cache)