When generating a wiki for our codebase, the resulting business logic summaries are very shallow:
• The codebase contains 20+ modules, but only 3–5 modules are included in the wiki output.
• The summarized content per module is minimal and not very useful.
• I am unsure whether this is due to incorrect usage or limitations of the model.
• Model used: openai/gpt-5.
Any guidance on improving coverage and summary quality would be appreciated.