chore: refactoring of the codebase is needed

[masterchief164]

No description provided.

[adamjonas]

Can you describe what you believe needs to be done ?

[masterchief164]

The inclusion of multiple transcribers with many options like diarization, summary creation, and topic detection has made the code a lot bloated. There are a lot of if else ladders that can be simplified. Also, the methods do not follow the SOLID principles. I'd like to address those issues and thus make the code a bit more maintanable.

[SarcasticNastik]

Refactor Strategy

Here, I outline a general idea for refactoring the codebase using an object-oriented approach.

Application Workflow

Pre-process source files

• Given source: Video|Audio|Yt-link, conditionally pre-process ( download and convert) the source files.

Current approach for detecting and specifying the source type and the corresponding processing is manual using source_type.
Automatic detection of file type and further processing should be achieved using either:

• Pattern matching using match (supported by python versions $\ge$ 10).
• Conditional matching.

Transcription

• Transcribe the audio input to text using either deepgram or whisper.

The corresponding codebase can be easily ported.

Post-process generated files

• Store the generated transcription from the previous section.
• Optionally, generate a PR for the transcription.
• Optionally, upload model outputs to AWS S3.

Again, all the corresponding functionalities can be easily ported.

App class

Methods

Corresponding to the previous section.

1. pre_process (needs a better name)

   • detect_source_type
   • download (conditionally)
   • convert (conditionally)

2. transcribe

3. post_process (needs a better name)

   • write_transcription_to_md
   • create_pr (conditionally)
   • upload_to_s3 (conditionally)

4. process: Complete workflow. (end-user API)

[kouloumos]

Codebase has been refactored using an object-oriented approach into a structured four-stage process. Most of the work has been done with #118 and 2764a7f, with additional changes in miscellaneous commits that can be found in the git history.