https://vapi.ai/ logo
Join Discord
Powered by
Urgent: Assistant Not Repeating Email Address
# support
s
Hi @Vapi @Vapi Support Bot @Shubham Bajaj , I am facing an issue where the assistant fails to repeat the email address after the user provides it. For reference, in call ID 
d5b8ba68-917b-4162-a7e9-f90d1ca593c1
, you can clearly hear the issue around the 2:30 timestamp— the assistant simply omits the email when trying to repeat it back. This is critical for our workflow, and we’d appreciate your urgent support in resolving it.
v
Hi there! 👋 I understand you're facing an urgent issue with your assistant not repeating email addresses. While I can't access the specific call recording you mentioned, I can definitely help with information about how this should work and potential solutions to fix it. ## How Email Repetition Should Work According to the [documentation](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1331353180570914878), the Vapi assistant should: - Guide users to spell out their email address one character at a time - Use "at" for "@" and "dot" for "." - Repeat the email back for confirmation: "I heard 'g-o-l-d.' Is that correct? Please confirm or correct me" ## Debugging Steps Here are some steps you can take to debug this issue: 1. **Check Assistant Configuration**: Review your assistant's configuration to ensure it's correctly set up to handle and repeat user input, specifically email addresses. 2. **Review Logs**: Check logs for the calls where the issue is occurring to see if there are any error messages or patterns. 3. **Verify Webhook Configuration**: Ensure that your webhook is properly configured if you're using one. 4. **Check the Assistant ID**: Make sure the assistant ID is correctly set on your phone number. ## Possible Workarounds Several users have found workarounds for similar issues: 1. **Use Webhooks**: One user reported: "I used the webhook to response then it is correct, rather than using Vapi module" according to [this Discord thread](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1334661059298066514). 2. **Adjust System Prompt**: Modify the 
systemPrompt
in your assistant configuration to explicitly instruct the assistant to repeat email addresses. 3. **Try a Different LLM**: Switching to a different LLM (like 4o-mini) has resolved similar issues for some users. 4. **Enable SSML Parsing**: This might give you more control over how the assistant handles and repeats the email address.
Would you like me to provide more details on any of these approaches or help with other aspects of troubleshooting this issue? Sources: - [Discord Thread 1331353180570914878](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1331353180570914878) - [Discord Thread 1334661059298066514](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1334661059298066514) - [Vapi Documentation](https://docs.vapi.ai/examples/voice-widget)
s
@Shubham Bajaj help me out please its little urgent here we are presenting our product in bahrain city summit tommorow
k
looking into it.
🔵 10:13:58:296 Voice Input Formatted: " Just to confirm, the email address you've provided is:. Please confirm if this is correct so I can proceed with the next steps.", Original: " Just to confirm, the email address you've provided is: [brian.pest@axlai-pest.accelnomics.com](brian.pest@axlai-pest.accelnomics.com) Please confirm if this is correct so I can proceed with the next steps."
Hey Sai, What actually happened here is any content in between the markdown symbols or syntax was trimmed off because of the default formatting plan. Could you toggle off all of the markdown related utility functions for your voice assistant voice? Basically, you have to update your assistant settings through API under the formatting plan which is under chunk plan of the assistant and disable all of the markdown related formatting utilities.
Copy code
curl -X PATCH "https://api.vapi.ai/assistant/{ASSISTANT_ID}" \
  -H "Authorization: Bearer {YOUR_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
  "voice": {
    "chunkPlan": {
      "enabled": true,
      "formatPlan": {
        "enabled": true,
        "formattersEnabled": [
          "newline",
          "colon",
          "acronym",
          "dollarAmount",
          "email",
          "date",
          "time",
          "distance",
          "unit",
          "percentage",
          "phoneNumber",
          "number"
        ]
      }
    }
  }
}'
Let me know if you require.
s
Hi @Shubham Bajaj First of all the email is taken wrong i tried my best to pronounce it letter by letter but still i couldn't make the assistant take the correct email, the actual email is 
brian.test@axlai-test.accelnomics.com
Can you please suggest some way to achieve this. coming to what you suggested to turn off the markdown related utility functions Can you please elaborate how can i do that using web sdk i am just doing vapi.start(assistantOptions) and this is how the assistantOptions look like this confirm me where should i add what to make the markdown turn off here { name: botName, firstMessage: 
Hello ${user.firstName}, I'm ${botName}, your dedicated ${roleName}. How can I assist you today?
, transcriber: { provider: "deepgram", model: "nova-2", language: "en-US", }, voice: { provider: "playht", voiceId: voice, }, model: { provider: "openai", model: "gpt-4o-mini", messages: [some messages here], tools: [some tools here], }, }
k
Can you send me a latest call-id?
s
Hi Sahil I already mentioned the call ID 
d5b8ba68-917b-4162-a7e9-f90d1ca593c1
I am again briefing my issues Issue 1: Despite of trying hard (spelling letter by letter) I am not able to make the assistant take the correct email. what i want : Suggest me some good ways to achieve this so that the assistant takes correct email Issue 2: Once the email is taken its not pronouncing it back maybe due to some marked down related issues as noticed by Shubham what i want : if it is due to markdown i have shared the assistantOptions above how am i passing what changes should i do to to make turn off markdown please tell me
k
for issue 1, i would recommend using word pairs with the letters to increase accuracy. i would test that first to see if that is even causing issue 2. let us know if it works any better.
s
hey kyle could you please elaborate with an example what does using word pairs with the letters means ?and please also check and update me for the second issue please
k
For word pairs, you could have the agent repeat the letter with a word like "E as in Echo" to further confirm the correct spelling. For the second issue: - ✅ Do **\*\* NOT \*\*** request clarification unless explicitly indicated. Remove the Asterisks/Bold formatting in the prompt to prevent the word "not" from being accidentally excluded in the instructions
k
@Sahil
PreviousNext
Powered by