●Change8

1.4.0

Breaking Changes
📦 crewaiView on GitHub →
⚠ 2 breaking✨ 6 features🐛 7 fixes⚡ 1 deprecations🔧 6 symbols

Summary

This release introduces first-class MCP support and LLM message interceptor hooks, while refactoring embedding configurations and fixing critical flow state and caching issues.

⚠️ Breaking Changes

  • The 'embedder' parameter has been renamed to 'embedding_model' across tool documentation and configurations.
  • VectorDB is now a required parameter across tool documentation.

Migration Steps

  1. Update code references from 'embedder' to 'embedding_model' in tool configurations.
  2. Ensure 'vectordb' is explicitly provided in tool implementations where it was previously optional or omitted.

✨ New Features

  • Add support for non-AST plot routes
  • Implement first-class support for Model Context Protocol (MCP)
  • Add Pydantic validation dunder to BaseInterceptor
  • Add support for LLM message interceptor hooks
  • Cache i18n prompts for efficient use
  • Enhance QdrantVectorSearchTool

🐛 Bug Fixes

  • Fix issues with keeping stopwords updated
  • Resolve unpickleable values in flow state
  • Ensure lite agents course-correct on validation errors
  • Fix callback argument hashing to ensure caching works
  • Allow adding RAG source content from valid URLs
  • Make plot node selection smoother
  • Fix duplicating document IDs for knowledge

🔧 Affected Symbols

BaseInterceptorQdrantVectorSearchToolMCPembedderembedding_modelvectordb

⚡ Deprecations

  • The 'embedder' parameter is deprecated in favor of 'embedding_model'.