crewAI shipped version 1.15.28, adding a 'crewai eval' command that reports where a stopped model broke and adopting the LiteLLM catalog for context windows. The release bundles numerous bug fixes around flows, deployments, and provider integrations including Anthropic, OpenAI GPT-6, and RAG.
crewAI published version 1.15.29 on GitHub, adding LiteLLM-compatible kwargs, Azure DocumentDB vector search, and expanding crewAI eval to run crews and export evaluations. The release also promotes AgentExecutor out of experimental status and fixes bugs across memory, tool calls, and LanceDB storage.
crewAI published version 1.15.29 on GitHub, adding LiteLLM-compatible kwargs, Azure DocumentDB vector search, and expanding crewAI eval to run crews and export evaluations. The release also promotes AgentExecutor out of experimental status and fixes bugs across memory, tool calls, and LanceDB storage.
crewAI shipped version 1.15.34 on GitHub, adding a standalone LiteLLM catalog, a crewai skill search command, crewai run --inputs support for Python flows, and deploy create --zip. The release also packs roughly 30 bug fixes covering LiteLLM streaming, structured output recovery, API key validation, and Python 3.10-3.12 compatibility, plus migration notes for the removed 1.15.29 agent executor.
crewAI shipped version 1.15.34 on GitHub, adding a standalone LiteLLM catalog, a crewai skill search command, crewai run --inputs support for Python flows, and deploy create --zip. The release also packs roughly 30 bug fixes covering LiteLLM streaming, structured output recovery, API key validation, and Python 3.10-3.12 compatibility, plus migration notes for the removed 1.15.29 agent executor.
crewAI shipped version 1.15.27, adding XPU device support for the OpenCLIP provider, DeepInfra as an OpenAI-compatible provider, and per-run cost and time tracking in crewai eval --models. Bug fixes address stopSequences sent to OpenAI models, GitHub loader source attribution, and SQLite flow initialization contention.
crewAI shipped version 1.15.27, adding XPU device support for the OpenCLIP provider, DeepInfra as an OpenAI-compatible provider, and per-run cost and time tracking in crewai eval --models. Bug fixes address stopSequences sent to OpenAI models, GitHub loader source attribution, and SQLite flow initialization contention.
crewAI shipped version 1.15.18 on GitHub, promoting conversational flows to stable alongside deployment tracking, router response-format declarations, and several bug fixes. The stable promotion signals the conversational flow API is ready for production use by agent developers.
crewAI shipped version 1.15.18 on GitHub, promoting conversational flows to stable alongside deployment tracking, router response-format declarations, and several bug fixes. The stable promotion signals the conversational flow API is ready for production use by agent developers.
crewAI shipped three rapid releases on GitHub: 1.15.19 (September 4), 1.15.20 (September 4), and 1.15.21 (September 9). The updates add a Clipper integrations client, route DashScope models through the native provider, fix a gpt-4o-mini context-window error, and bump pypdf and nltk for security vulnerabilities.
crewAI shipped three rapid releases on GitHub: 1.15.19 (September 4), 1.15.20 (September 4), and 1.15.21 (September 9). The updates add a Clipper integrations client, route DashScope models through the native provider, fix a gpt-4o-mini context-window error, and bump pypdf and nltk for security vulnerabilities.