TreeInstruct provides a framework to build hierarchical, decision-tree based prompting pipelines for large language models. Users can define nodes representing prompts or function calls, set conditional branches based on model output, and execute the tree to guide complex workflows. It supports integration with OpenAI and other LLM providers, offering logging, error handling, and customizable node parameters to ensure transparency and flexibility in multi-turn interactions.

Whiz is designed to provide a robust foundation for developing intelligent agents that can perform complex conversational and task-oriented workflows. Using Whiz, developers define "tools"—Python functions or external APIs—that the agent can invoke when processing user queries. A built-in memory module captures and retrieves conversation context, enabling coherent multi-turn interactions. A dynamic planning engine decomposes goals into actionable steps, while a flexible interface allows injecting custom policies, tool registries, and memory backends. Whiz supports embedding-based semantic search to fetch relevant documents, logging for auditability, and asynchronous execution for scaling. Fully open-source, Whiz can be deployed anywhere Python runs, enabling rapid prototyping of customer support bots, data analysis assistants, or specialized domain agents with minimal boilerplate.