The recommended method for working with audio data and LLMs is AssemblyAI's LeMUR: https://www.assemblyai.com/blog/lemur-early-access/
This folder is useful in the event that you want to embed your audio data by semantically related chapters or 'sections' as opposed to using arbitrary character splitting.
-
assembly_chapters.py- example of transcribing an audio file with auto_chapters enabled. -
chapter_helpers.py- splits the transcript by chapter and includes the full text of each chapter in each section. -
assembly_embed_by_chapter.py- use chroma + openai embeddings to embed each section of the transcript and do similarity search