New Features

• Public Search Mode: added unauthenticated read-only access to global search and previews while keeping write operations protected.
• Documentation Datasets: ships metadata and tooling to import reference documentation (Go, PowerShell, Python, Rust, Node.js, MDN).
• StackExchange Extractor: replaced the StackOverflow extractor with a generic StackExchange extractor indexing answers with metadata.
• Local File Previews: local Markdown and Org files now render as HTML previews with automatic title extraction and clean syntax.
• MCP History & Capabilities: added get_history tool to the MCP endpoint, date filters support, and semantic search capability reporting.
• Browser Extension Upgrades: added configurable hotkeys for indexing controls and badge indicators showing if a page is already indexed.
• Import & Export Improvements: supports HTML files, multiple input files, compressed JSON, date-range filtering, and directory imports.
• Search Sorting & JSON Output: added sorting by newest date, domain, or most visited pages, and a format=json parameter to the search endpoint.
• History RSS Feed: the history endpoint and UI now expose an RSS feed for recently indexed documents.

Enhancements

• File indexing startup runs in the background to avoid blocking server start.
• Watched directory indexing now uses a proper, non-duplicating queue.
• Added CLI indexing flag to explicitly allow sensitive content when requested.
• Semantic search concurrency limits and merging of shared and user-specific vector results.
• Redesigned history timeline displaying indexed version counts and day-unloaded item indicators.
• Refined web UI layout, dark theme, search input, and configurable web title/subtitle branding.
• Added type facets and optimized rendering to hide inactive facet categories with single candidates.
• Built-in yt-dlp is now pre-installed in the official Docker image.

Bug Fixes

• Fixed Lobsters extraction to match updated HTML structure.
• Fixed document label preservation and batch user ID handling.
• Fixed text wrapping and layout squeeze on history and search views.
• Improved Base URL consistency for CSRF checks and invalid configurations.
• Recovered PDF parsing panics and addressed memory issues in PDF dependencies.
• Prevented reindexing from stopping entirely when a single document extractor aborts.
• Fixed database migration side-effects on UpdatedAt timestamps.