Seeking Advice on generating documentation using LLMs

I’m working on documenting a web application that involves a complex structure with numerous classes. My goal is to use Large Language Models (Azure OpenAI) to generate scenario-based documentation with detailed use cases. However, I’m facing challenges in managing the input for gpt-35-turbo, especially given the large number of classes involved.

Could you share your best practices or strategies for:

1. Handling large codebases with multiple classes when using LLMs.
2. Efficiently managing context limits in LLMs to maintain relevance.
3. Breaking down complex code structures into manageable chunks for documentation purposes.

Any advice, tools, or resources would be greatly appreciated!