Perf: Async operations in parallel wherever possible

[JayGhiya]

No description provided.

[vipinshreyaskumar]

Quantify minuscular work to parallelise , and parallelise.

[JayGhiya]

• Functions of a class can be parallelised inside codebasesummary.py (line number 68) - https://github.com/unoplat/unoplat-code-confluence/blob/main/unoplat-code-confluence/unoplat_code_confluence/summary_parser/codebase_summary.py . Then it has to be awaited until all functions of that particular class are finished. Then Classes of a particular package can be parallelised - (line number 72) . Then similar for package - line number 75

[JayGhiya]

Parallelism has to be controlled through semaphore in asyncio and externalised for user to make sure he does not hit limits on his llm provider.

[JayGhiya]

#90 we are also including dspy suggest in the same issue as perf is affecting comprehension

[JayGhiya]

testing with gpt4o-mini to see better performance for oss repos.

[JayGhiya]

Resuming this work and trying to assign customised parallelism at class level with n:n instances of llm provider.