Release Notes¶

Librovore 1.0a3 (2025-09-01)¶

Remove unreliable signature detection system to provide cleaner, more focused output with better performance.

CLI: Fix exception rendering to respect –display-format setting for consistent JSON output.
MCP: Fix content length issues that caused tool failures by implementing proper response size control through user-specified limits.

Add MkDocs inventory processor to support pure MkDocs sites using search_index.json files.
Implement parallel URL pattern probing for improved performance and network efficiency.
Transform JSON output format from dictionary-based to structured dataclass objects with enhanced error handling.

Add automatic URL pattern extension for documentation sites. The system now automatically discovers working documentation URLs when the base URL fails, significantly improving compatibility with ReadTheDocs, GitHub Pages, and other hosting platforms that use versioned paths like /en/latest/ or /latest/. Previously failing sites like requests.readthedocs.io and docs.pydantic.dev now work seamlessly with transparent URL resolution.
Improve error messages with structured, user-friendly responses. Replace generic “No processor found” errors with clear, actionable messages that distinguish between inventory and structure detection failures. Error responses now include specific titles, detailed explanations, and practical suggestions for resolving issues with unsupported documentation sites.

Fix JSON serialization error for frigid.Dictionary objects in CLI and MCP server responses. Previously, commands like summarize-inventory with --group-by parameter would fail with “Unable to serialize unknown type” errors. Enhanced serialization now properly handles all internal data structures for consistent JSON output.

No significant changes.