I configured an assistant and used it for 2-party conversation session, the use case is speaker speaks in Sanskrit, and the talk back from the tutor agent is also in Sanskrit. When I analyse the call logs (attached here), unable to clearly trace if the ASR (transcriber) has transcribed accurately or at what point there is issue in the overall flow with respect to translation accuracy from the logs ? Logs Give Below - { "role": "system", "time": 1746681806286, "message": "You will receive transcription in Sanskrit and you will respond with natural conversations in sanskrit, keep the conversation short and close out with thank you in Sanskrit as the final point in session.", "secondsFromStart": 0 } { "role": "bot", "time": 1746681806582, "source": "", "endTime": 1746681808450, "message": "Namo Namo", "duration": 1.8679999999999999, "secondsFromStart": 0 } { "role": "user", "time": 1746681806582, "endTime": 1746681813474, "message": "नमो नमहे समस्कुजम पढ़ितुम इच्छामी", "duration": 6.892, "secondsFromStart": 0 }

To help diagnose your Sanskrit transcription issue, could you provide more of the conversation log that shows the complete interaction? The current log only shows one user message and one bot response, which isn't enough to trace where transcription or translation errors might be occurring. Ideally, I'd need to see: - Several turns of conversation - Any expected translations/responses - Which transcription model you're using for Sanskrit This would help identify if the issue is with ASR accuracy, translation, or another part of the conversation flow. Sources: - [Methods for analyzing call logs to trace issues in voice AI applications](https://discord.com/channels/1211482211119796234/1334142600772456508) - [How to determine the accuracy of ASR transcription from call logs](https://discord.com/channels/1211482211119796234/1342583376707850250)

Attached the file consisting of logs in a multi turn conversation, i am using Talkscribe transcription model for transcriber and gpt for response processing logic and 11Labs for TTS, the expected transations are not very clear in the intended target language (Sanskrit) https://cdn.discordapp.com/attachments/1369910033680437258/1369911662869610586/LogAnalysis.docx?ex=681d9528&is=681c43a8&hm=22f40c69fd1b280fa6f5f827dbb62ed508a015dc5953aa8e7953bc54765505f1&

Hi, can you please provide a call id for this call so we may investigate this issue?

this is the call id - 3b26c3a6-fc01-4391-86d1-3c655adb9b51

Hey! To help track down this issue, could you share: - When exactly this happened (the timestamp) - What response you expected to get - What response you actually got instead This would really help us figure out what went wrong!