[2] AI transcript

[MrOrz]

Design doc:

• https://g0v.hackmd.io/@cofacts/rd/%2Fwkx286lmTDaFUpgRhnUawQ (ListArticle part)
• https://g0v.hackmd.io/LxNqtaRhQbqkSFMsLShAiQ#AI

Implementation

• Implements createTranscript(), which can be re-used in CreateMediaArticle in the future.
  • Uses google vision API to perform OCR for images
  • Uses OpenAI Whisper API to perform STT for audios and videos
  • Includes integration tests that really calls the API to ensure that it's correctly connected to third-party APIs
• Include ffmpeg in Docker & Github action CI to process video and audio files
  • extract audio to reduce size & normalizes format

[coveralls]

coverage: 88.348% (+0.2%) from 88.141% when pulling 34ed2cd39f15b3606de95619f720400a15c754b1 on ai-transcript into aa5e85f7a62bf8562d8aa92b39b9b6d3f1487871 on master.

[MrOrz]

Merge is blocked by LINE bot scoring fix described in 2023/8/2 meeting

[MrOrz]

Current dependencies on/for this PR:

• refactor-ai-resp
  • PR #309 👈
  • PR #315
    • PR #313

This comment was auto-generated by Graphite.