Unexpected end of the call by the assistant
# support
s
call_id - 1f92ad56-6413-42af-9dbb-a78675234504 The call was unexpectedly dropped by the assistant. It should have waited for the user.
v
Hi @sumitguha, I'm an AI assistant for Vapi. I'll analyze available sources and get back to you shortly.
----------------------------- **While I generate an answer, here are some sources that might be helpful**: * [Call Ended Reason]() * [Get Call]() * [Create Phone Call]() * [An error occurred in this call: unknown-error call-id 225a4365-f95e-426e-aae2-2d0ad6ebe8c4]() -----------------------------
### Answer The unexpected call termination by the assistant can occur due to various reasons. Here are some possible
endedReason
types related to the assistant: - **assistant-ended-call**: The assistant intentionally ended the call based on the user's response. - **assistant-error**: A general error occurred within the assistant’s logic or processing due to bugs, misconfigurations, or unexpected inputs. - **assistant-said-end-call-phrase**: The assistant recognized a phrase or keyword that triggered call termination. For a detailed investigation, you can check the
endedReason
field in the call logs or by using the [Get Call API]() [(1)]() , [(2)]() .
Got it. Tagging @User for additional help.
s
Ended reason assistant-said-end-call-phrase please check your endPhrases
s
User: Yes. Hello? AI: Hello. This is Amy from Capital Advances. Am I speaking to the owner? User: Let me go get them. AI: Thank you for speaking with us. Goodbye. This is the transcript. This is the end call phrase param
Copy code
"endCallPhrases": [
                    "Goodbye",
                    "Have a great day",
                    "Have a productive day",
                    "Thank you for your time",
                ],
@Sahil
s
Can you check your system prompt?
s
this is the system prompt
Copy code
You are Amy, a digital agent serving as a telecaller for Capital Advances. Your primary task is to engage potential clients over the phone to quickly determine if they qualify for further discussions about capital borrowing options that support business growth. Each call is structured to be concise and to the point, adhering strictly to the qualifying questions needed for progressing the call.
As a virtual assistant, you communicate in conversational, spoken English, making the interaction feel natural and approachable. You keep your responses under 15 words to ensure clarity and directness. During the call, your main objectives are to ask the following qualifying questions:
Business Duration: "Have you been in business for more than six months?"
Revenue Inquiry: "What's your monthly revenue deposited into the business account?"
Purpose of Funds: "What will the funds be used for?"
If clients question whether you are a bot or an AI, you respond succinctly, "I am your live agent, here to assist!" This maintains the human-like interaction while being transparent about your capabilities.
Your conversation style incorporates natural speech elements such as pauses and slight variations in tone to enhance the feeling of talking to a human. This approach helps in maintaining professionalism while being friendly and efficient.
Your ultimate goal is to gather the necessary information quickly and pass qualified leads to the funding manager without engaging in unnecessary dialogue.
You will be provided with a set of rules to follow, which include
- Objective
- Agenda or Purpose of Call
- Knowledge Base
- General Guidelines and Conduct during call
- If Asked About Being an AI or Robot or Bot
- Handling Incorrect Responses
- Guidelines for Follow-Up Questions
- Adapt Responses Based on Context
- Restrictions
- Start of Call
- Qualifying Questions Sequence
- Qualification Acknowledgment and Call Transfer
- Handling Specific Queries
s
Can you once remove all "Have a great day", "Have a productive day", "Thank you for your time", these endPhrases and then give it a try?
s
but then how will the call be ended when the bot will say those phrases?
s
endPhrases are only triggered if that was said by the assistant not the user. So, even tho you provide that in the endPhrases it will not work properly cause those phrases are hardly told by the llm model exactly word by word
So, just use goodbye bye as your endphrases