File Search and Function Calling: Stuck at Required_Action?

API
, , , ,
1

Hello, I am currently working on an assistant that reads in a PDF file of a using File Search and then have the API scrape it for important info, which is to be outputted into a very specific excel file format using function calling. Here is the start of the function:

‘’’
{
“name”: “create_excel”,
“description”: “Create an Excel file with extracted information for all tabs”,
“parameters”: {
“type”: “object”,
“properties”: {
“agreements”: {
“type”: “object”,
“description”: “Data for the Agreements tab (Tab 1)”,
“properties”: {
“Code”: {
“type”: “string”,
“description”: “Code”
},
“Name”: {
“type”: “string”
},
“Type”: {
“type”: “string”
},
“Sub_Type”: {
“type”: “string”
},
“Status_Group”: {
“type”: “string”
},
“Status”: {
“type”: “string”
},
“Responsible_Office”: {
“type”: “string”
‘’’

But I am running into an issue where my poll creates eternally stuck at “required_action.” This seems to be the same point at which the assistant in the playground requires user input like “Success: True” to continue, however I see it already has found the data I want. When I try to access the output with:

messages = list(client.beta.threads.messages.list(thread_id=thread.id, run_id=run.id))
messages = [] Empty!

But when I try to access the run from:
run = client.beta.threads.runs.create_and_poll( thread_id=thread.id, assistant_id=assistant.id)

The run has all of my data + instructions but I can see the:
required_action=RequiredAction(submit_tool_outputs=RequiredActionSubmitToolOutputs(tool_calls=[RequiredActionFunctionToolCall In the run, immedieatly followed by all of the data I need in the correct format!

How do I get past this and just get my data in the correct JSON format to be turned into an excel file? I already have that code written, I would just rather get the correct output from the assistant. When I attempt to get past the required_action, by typing “Success: ‘True’” into the assistant playground, it just gives me a random summary I didn’t ask for. The API is clearly calling the correct function (I have detailed instructions), I just cant get the correct output!

My specific excel template is very long (37 tabs) so I have been wary of just putting that really long JSON format into the instructions but could try that.

Any help is greatly appreciated, sorry for the long post.

2

you need to submit that output. see the last part here:

// Submit all tool outputs at once after collecting them in a list
    if (toolOutputs.length > 0) {
      run = await client.beta.threads.runs.submitToolOutputsAndPoll(
        thread.id,
        run.id,
        { tool_outputs: toolOutputs },
      );
      console.log("Tool outputs submitted successfully.");
    } else {
      console.log("No tool outputs to submit.");
    }

however, please note that the final output in the message will be determined by your instruction.