May 5: Support for Jina and Pongo rerankers, Microsoft Teams feed, new YouTube downloader, bug fixes

New Features

  • Given changes in YouTube video player HTML, we have rewritten the YouTube downloader to support the new page format.

  • Added better handling of HTTP errors when validating URIs. Previously some websites were returning HTTP 403 (Forbidden) errors when validating their URI, or downloading content. Now Graphlit is able to scrape these sites, which previously returned errors.

  • Added support for updating content metadata in updateContent mutation. Now the video, audio, document, etc. metadata can be updated after the content workflow has finished.

  • Added query_contents_graph (and queryContentsGraph) functions to SDKs, which can be used to return nodes and edges from knowledge graph for visualization.

Bugs Fixed

  • GPLA-2544: Page relevance not filled-in in all situations

  • GPLA-2546: Not extracting links from PDF with Azure AI Doc Intelligence

  • GPLA-2557: Sporadically returning HTTP 500 from GraphQL API

  • GPLA-2573: Failed to re-ingest content which was deleted immediately after initial ingestion

  • GPLA-2575: Not validating for empty (non-null) parameters in mutations

  • GPLA-2578: Need to handle invalid JSON from LLMs; improper escaping or formatting

  • GPLA-2585: Failed to ingest encoded file with colon (:) in name

Last updated