1.10.0
📦 crewaiView on GitHub →
✨ 10 features🐛 13 fixes🔧 3 symbols
Summary
This release introduces significant feature enhancements, including improved tool resolution, JSON argument parsing, and user input handling in Flows. Numerous bug fixes address issues related to tool validation, schema handling, and flow persistence.
Migration Steps
- If you were using the CLI HTTP client based on requests, note that it has been migrated to httpx.
- Review documentation for litellm dependency notes if using non-native LLM providers.
✨ New Features
- Enhanced MCP tool resolution and related events
- Updated lancedb version and added lance-namespace packages
- Enhanced JSON argument parsing and validation in CrewAgentExecutor and BaseTool
- Migrated CLI HTTP client from requests to httpx
- Added versioned documentation
- Added yanked detection for version notes
- Implemented user input handling in Flows
- Enhanced HITL self-loop functionality in human feedback integration tests
- Added started_event_id and set in eventbus
- Auto updated tools.specs
🐛 Bug Fixes
- Validated tool kwargs even when empty to prevent cryptic TypeError
- Preserved null types in tool parameter schemas for LLM
- Mapped output_pydantic/output_json to native structured output
- Ensured callbacks are ran/awaited if promise
- Captured method name in exception context
- Preserved enum type in router result; improved types
- Fixed cyclic flows silently breaking when persistence ID is passed in inputs
- Corrected CLI flag format from --skip-provider to --skip_provider
- Ensured OpenAI tool call stream is finalized
- Resolved complex schema $ref pointers in MCP tools
- Enforced additionalProperties=false in schemas
- Rejected reserved script names for crew folders
- Resolved race condition in guardrail event emission test