2.98.7

2.98.7

Fixed

• Feed refreshes no longer force a full download when processed episodes are intentionally hidden by title rules.
• Anthropic models that reject forced tool calls use compatible structured output. Per-stage reasoning settings support adaptive effort while preserving token budgets for older models. (#827)
• Anthropic calls cache static system prompts for 5 minutes. Usage totals include cached input, and cost estimates account for cache reads and writes. (#828)

2.98.6

Fixed

• Notification and webhook controls meet the 44px mobile touch target.

2.98.5

Fixed

• Filters across all Patterns tabs use aligned, equal-width controls. The Notifications section groups Email and Webhooks into collapsible subsections. Podping server settings appear only while Podping is enabled and use compact node controls.
• The Settings page and documentation use Configuration Import / Export for runtime configuration backup and restore.

2.98.4

Added

• When a webhook template fails, MinusPod uses a readable text body for text/* endpoints and JSON otherwise; test results identify each affected event. (#826)

Fixed

• Email notifications and Webhook notifications collapse independently. Webhook cards display full URLs and keep event labels on one line.
• Webhook test summaries show template fallbacks separately from delivery status and wrap on narrow screens. Text fallbacks collapse whitespace, and dispatch error logs redact URL credentials and exception details.
• The ntfy template example handles error messages on both failed episodes and alert events.
• Regenerating chapters records a separate processing-history entry and episode log, including failed attempts, without changing the previous audio run's results or token totals.

2.98.3

Added

• Pattern Cleanup can suggest category changes alongside text and sponsor edits, including separate categories for split patterns.
• Podping settings let users add, edit, remove, and reset RPC servers through the UI and API.
• Export and restore runtime configuration for global settings, all feeds, or selected feeds. Preview configuration changes before applying them; episodes, history, and unselected feeds are preserved. (#823)

Fixed

• Retired search-index cleanup uses smaller write batches so it releases the database lock more often.
• Processing history shows chapter-generation warnings. Failed chapter regeneration keeps existing chapters and reports the failure. Chapter generation respects cancellation and provider changes, and detailed response text is logged only at debug level.
• Changing a provider, endpoint, or stage route clears affected model selections and shows a reminder to choose a model. Changing an API key keeps model selections. API requests can set replacement models in the same update.