VAPI not fetching proper office information for a ...
# support
d
I gave the following information to assistant. When I ask for tomorrow's business hours, bot asked for the location name, and it said the location is closed. Please let me know how to fix it. Call ID: 6a7d7189-7460-49db-8c43-052077b7f0b8, f9e6a719-dd43-41dc-bd70-33dc1f51e3fe Assistant information ----------------------- The current date is {{"now" | date: "%A, %B %d, %Y", "America/New_York"}} and the time is {{"now" | date: "%I:%M %p", "America/New_York"}}. Business Hours: Monday: 9 to 2 PM. Tuesday: Closed. Wednesday: 9 to 6 PM, Thursday: Closed, Friday 9 to 4.30 PM, Saturday: Closed, Sunday: Closed. Procedure for Providing business or office or working or opening hours: 1. Day Handling - If the user provides a specific day (e.g., "Monday"), use that day. - If no day is mentioned, ask: "Which day would you like the office hours for?" 2. Location Handling - If the user provides a location name, use it. - If no location is provided, prompt: "Can you tell me the location you’re asking about?" - Note: Location is required to proceed. 3. Retrieving Hours - Use the provided location and day to look up the business hours from the available data.
v
To help you further, could you please confirm whether you are providing a location name when asking for tomorrow's business hours, and if so, what exact input are you giving to the assistant? This will help determine if the issue is with location handling or day matching in your setup. Source: - [Vapi Personalization Documentation](https://docs.vapi.ai/assistants/personalization)
d
Yes. I provided the heading as Yonkers Location: Yonkers Location: Address: xxxxxxxx, Phone: xxxxxxxx, Fax: xxxxxxxxxx, Business Hours: xxxxxxx Another Call ID: 5562eb42-f3ad-4d6c-af5f-0149259e7b01
k
Looking into it
k
It sounds like the issue is with how the prompts are set up, which might not be giving you the results you want. Try checking out models like 4.0, 4.1, or those designed for research help. Share your transcript, system prompt, what you hoped to get, and what you actually got with these models.
Thinking through the results more deeply can help you spot where things are going off track. Keep tweaking your system prompts and running the chat again. Each time, you'll get closer to matching your inputs with the desired outcomes. With some adjustments and persistence, you’ll likely find the right approach.
This is a prompting nuance problem, so it has to be resolved by you. Let me know how it goes for you.