AHi @Alex, I'm an AI assistant for Vapi. I'll analyze available sources and get back to you shortly.
A-----------------------------
While I generate an answer, here are some sources that might be helpful:
While I generate an answer, here are some sources that might be helpful:
- Create Phone Call
- Create Assistant
- List Calls
- Update Phone Number
-----------------------------
AAnswer
To address the issue you're encountering with custom functions not being properly called during outbound calls via the API, while working fine from the dashboard, it's important to ensure that your JSON structure for the API call includes all necessary details for custom function execution. Unfortunately, without seeing the exact JSON structure you're using, I can only provide general advice based on the official documentation.Ensure your JSON payload for the 'Create Phone Call' API:
- Correctly specifies the
assistantIdorassistantobject that includes the definition of custom functions. This is crucial for the assistant to know which functions are available for calling during the conversation (1) . - Includes the
serverUrlproperty within the assistant object if you're defining a transient assistant directly in the call payload. This URL is where Vapi will send requests to trigger custom functions (2) . - Has valid authorization headers and uses the correct HTTP method and endpoint URL as per the 'Create Phone Call' API documentation (1) .
- Make sure the
functionsarray within your assistant definition includes the correct
,name
status (if applicable), andasyncserverUrlfor each function (2) . - Ensure your server endpoint, specified in the
serverUrl, is correctly processing incoming requests from Vapi and responding as expected. The server must handle function calls by returning a JSON object with the expected output for the assistant to use (3) .
Adocs.vapi.ai/server-url/events) can also provide insights on how your server should respond to different types of requests from Vapi, including function calls.
If you've verified all the above and the issue persists, it might be related to the specifics of how the JSON payload is structured or how the server processes the requests. Without the actual JSON structure, these are the most relevant points to check based on the information provided in the official documentation.
For more detailed assistance, you might want to compare your JSON structure with examples provided in the documentation, such as the outbound sales example that includes a detailed assistant definition and server URL configuration (4) .
If you've verified all the above and the issue persists, it might be related to the specifics of how the JSON payload is structured or how the server processes the requests. Without the actual JSON structure, these are the most relevant points to check based on the information provided in the official documentation.
For more detailed assistance, you might want to compare your JSON structure with examples provided in the documentation, such as the outbound sales example that includes a detailed assistant definition and server URL configuration (4) .
WThe authorization headers, http method, and endpoint url are correct, server url is correct and working too. Can you please check my assistant object to see if there's any mistakes? @Ask Inkeep
AThanks for the follow-up, @Alex. I'll get back to you soon.
A-----------------------------
While I generate an answer, here are some sources that might be helpful:
While I generate an answer, here are some sources that might be helpful:
- Update Assistant
- Create Assistant
- Function Calling
- List Assistants
-----------------------------
AAnswer
Sorry, I wasn't able to find any information about your question. You can get more help at Contact Us
TYou need to add descriptions to the function and its parameters
Wyea I did
Wjust shortened them here for brevity because I wanted t ask the AI support here
Wdescriptions and parameters an system prompt are there actually in my real json data
Wjust want to check if there's anything wrong with my json structure, especially the one for assistant objectr
T3.5 is ass when it comes to function calling
TTry 4
Wbecause I am using the exact same 4 functions and exact same server urls, and they work fine in dashboard
WI tried in dashboard using 3.5 too, works without issue, tried countless times
Wand in the api call, I have tried 4 as well, didnt really work somehow
Wso the llm is probably not the cause? I'm thinking might be my json
W@A. Salem just tested again 10 seconds ago with 4turbo0125, same issue
W@Ask Inkeep The authorization headers, http method, and endpoint url are correct, server url is correct and working too. Can you please check my assistant object to see if there's any mistakes?
AThanks for the follow-up, @Alex. I'll get back to you soon.
A-----------------------------
While I generate an answer, here are some sources that might be helpful:
While I generate an answer, here are some sources that might be helpful:
- List Assistants
- Create Assistant
- Update Assistant
- Get Call
-----------------------------
AAnswer
Sorry, I wasn't able to find any information about your question.
AGot it. Tagging @Vapi Team for additional help.
Lcan you send some call_id as well as your assistant id
