The library also provides the `@prompt_chain` decorator for creating complex logic by mixing LLM queries and function calling with regular Python code. It also supports the use of the `StreamedStr` class to process text while it is being generated, and the `Iterable` return type annotation for processing each item while the next one is being generated. The library can be configured using environment variables or arguments passed when initializing an instance in Python.
Key takeaways:
- 'magentic' is a Python library that allows easy integration of Large Language Models (LLMs) into Python code. It provides a compact, atomic, transparent, compatible, and type-annotated way to query LLMs.
- The '@prompt' decorator is used to define a template for an LLM prompt as a Python function. The arguments are inserted into the template, and the prompt is sent to an LLM which generates the function output.
- The '@prompt_chain' decorator resolves 'FunctionCall' objects automatically and passes the output back to the LLM until the final answer is reached. This allows for increasingly complex LLM-powered functionality.
- The 'StreamedStr' class can be used to stream the output of the LLM, allowing the text to be processed while it is being generated. Structured outputs can also be streamed from the LLM using the return type annotation 'Iterable'.