May 11: Support for Amazon Bedrock models, McPoogle MCP search engine, bug fixes

New Features

  • πŸ’‘ Graphlit now supports Amazon Bedrock models for LLM completion, preparation and extraction. Currently, we only support 'bring your own key' for models in your Bedrock environment, but will support per-credit usage of Bedrock models in an upcoming release.

  • We have launched our MCP search engine, called McPoogle, for finding MCP servers and tools. We also launched McPoogle as a remote MCP server, which you can use within AI-native IDEs such as Cline, Cursor or Windsurf.

  • We have launched a remote MCP server for 'Ask Graphlit', where you can get code generation of Graphlit SDK samples within AI-native IDEs such as Cline, Cursor or Windsurf.

  • We have added support for providing a 'seed' image when performing image generation. You can select an example image content, by reference, and use that to guide the image generation model for following your user prompt.

Bugs Fixed

  • GPLA-4194: Parsing raw GitHub README has extra HTML at end

  • GPLA-4202: Round-trip markdown to TextDocument loses Markdown list items

  • GPLA-4235: Hybrid RRF threshold is removing too many sources

  • GPLA-4233: Can't send just image & system prompt to Anthropic LLM

Last updated