I have a well-working custom GPT that calls the Airtable API to pull a table, and find relatively matching records. Currently trying to migrate this to an Assistant callable via API.
Right now when I pass the stringified Airtable API response in:
const run = await openai.beta.threads.runs.submitToolOutputs(
retrievedRun.thread_id,
retrievedRun.id,
{
tool_outputs: [
{
tool_call_id: retrievedRun.required_action.submit_tool_outputs.tool_calls[0].id,
output: JSON.stringify(airtable_data),
},
],
}
);
I receive the message “Content not found or not text type.” Worth noting, when I shorten it to just an item or two, it figures it out. So it doesn’t seem like an actual response length issue, maybe a comprehension issue?
But as far as I’m aware, unlike the GPT API relationships, for the Assistant functions I don’t go as far as defining the response structure to help it?
Please let me know if you have any ideas.