Inconsistencies between logs and transcript
Godefroy
I excepted to find raw LLM input/output in the transcript, but it seems to be processed with something else. Transcript is incoherent (especially in non-english language) and is missing parts.
For example, on this screenshot, you'll the AI answered "Est ce la rapidité de développement," (voice input), and becomes "et seule la rapidité de développement " in the transcript, which no sense.