v2.27.0-rc1

deepset-ai has launched Haystack v2.27.0-rc1, a significant pre-release update to its popular open-source framework for building production-ready RAG (retrieval-augmented generation) and LLM applications. The headline feature is "Smart Pipeline Connections," which introduces automatic implicit list joining. This allows pipeline components that expect a list input to automatically combine multiple inputs of compatible types—such as merging a plain query string with a list of ChatMessage objects into a single list[ChatMessage]—without requiring developers to write extra conversion or joining components. This simplification enables more flexible and cleaner pipeline architectures.

The release also delivers major quality-of-life improvements for developers working with DocumentStores. The InMemoryDocumentStore now includes metadata inspection and filtering utilities previously only available in persistent stores, like `count_documents_by_filter` and `get_metadata_field_min_max`. This allows for full local prototyping and debugging with the same logic used in production. Additionally, the AzureOpenAIChatGenerator now exposes a `SUPPORTED_MODELS` variable listing compatible models, including OpenAI's newest offerings like `gpt-5-mini` and `gpt-4o`. A security fix also sanitizes template variables in ChatPromptBuilder to prevent crafted inputs from being interpreted as structured content.