Release Date: July 2025
Overall Status: Production Ready
Breaking Changes: None
HyperFlow introduces a revolutionary approach to service configuration with the new Service Selection node. This node creates first-class service configuration objects that transform static service settings into flowing data values. Service configurations can now be stored in Key-value stores, modified by transform nodes, and selected dynamically by LLM decision-making. This enables powerful new workflow patterns including dynamic service selection, configuration management, and comparative service testing across the entire HyperFlow service ecosystem.
The Service Selection node supports all HyperFlow service types from LLM generation to content processing to knowledge database operations, making service configurations reusable across workflows and enabling sophisticated automation where service choices depend on runtime conditions.
HyperFlow now provides enterprise-grade access to OpenAI models through Microsoft Azure, combining OpenAI's AI capabilities with Azure's enterprise features. This integration supports both chat and embedding models with dual authentication options (API key or Microsoft Entra ID), enhanced security through private endpoints, and regional deployment controls for organizations with strict data governance requirements.
Azure OpenAI enables enhanced security, compliance, and regional deployment options while maintaining full compatibility with existing OpenAI-based workflows.
A powerful new multi-input/multi-output Python processing node designed for advanced data routing and transformation workflows. Unlike template-based transform nodes, Python Transform provides a pure Python execution environment with 3 configurable inputs, 3 independent outputs, and dual-exit flow control for sophisticated conditional branching and stateful programming patterns.
The node enables complex data flow patterns, feedback loops for iterative processing, and advanced RAG applications with its multi-input/output architecture and enhanced Python environment featuring special globals like keyValueStore and SafeSQLExecutor.
A powerful new data manipulation node that extracts and transforms data from any input type using JSONata expressions. The node accepts universal input via an "any" input type and provides flexible output formatting (Text, JSON, Number, Boolean) with dynamic type validation. This enables sophisticated data manipulation including field extraction, structure transformation, and calculations on complex data without requiring custom Python code.
LLMs in HyperFlow can now store and retrieve persistent information across conversations through automatically available memory tools. When enabled on Call LLM nodes, the system provides memory management functions (store_memory, retrieve_memory, list_memory_keys, delete_memory) that leverage the HyperFlow Key-value store with entity isolation for per-user memory separation.
This feature enables chatbots and AI assistants to remember user preferences, past interactions, and important context between sessions, creating more personalized and contextually aware AI experiences.
A new iterative processing node that enables systematic testing and comparison by cycling through multiple input connections sequentially. The node supports all major data types including configured services, making it ideal for A/B testing different service configurations, processing multiple data sets systematically, or implementing automated comparison workflows.