strata

This capability allows users to define and call functions using a schema-based approach, which standardizes how functions are registered and invoked across multiple AI model providers. By leveraging a unified protocol, it enables seamless integration with various backend models like OpenAI and Anthropic, ensuring that function calls are handled consistently regardless of the underlying model. This design choice enhances interoperability and reduces the complexity of managing different API specifications.

This capability provides a mechanism for managing the state and context of interactions with AI models, allowing for more coherent and contextually aware responses. It employs a context management system that tracks user inputs and model outputs over a session, enabling the system to maintain continuity in conversations or tasks. This approach ensures that the AI can reference previous interactions, improving user experience and relevance of responses.

This capability enables users to orchestrate workflows that involve multiple AI models, allowing for complex task execution that can leverage the strengths of different models. It uses a pipeline architecture where tasks can be defined in a sequence, with outputs from one model serving as inputs to another. This design allows for sophisticated processing chains, making it suitable for applications that require combining different AI functionalities.