Structured data analysis
# support
t
Call id 1: f83263f5-b921-4780-bbb2-7bed1ec259c0 Call id 2: e530a528-9dc6-4dfe-8e59-9b02a645781f For these 2 call ids, I did not receive the structured data in the structuredDataPlan. But I can see the tokens are consumed, but I am not seeing the output. Could you please help me with this?
v
Hi there! 👋 I don't have specific information about those exact call IDs you mentioned. However, I can definitely help you understand why tokens might be consumed without generating structured data output in your structuredDataPlan. Based on the information I've found, there are several potential reasons why this might be happening: ### Schema or Configuration Issues Your structured data schema might be incorrectly defined or have inconsistencies. According to [troubleshooting information](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1265637856383078461), it's recommended to "verify that the structured data schema is correctly defined and has not been altered." One specific issue to check is whether you've removed or renamed properties in the Dashboard without updating the required fields. The analyzer can get confused if "`analysisPlan.structuredDataSchema.properties.owner_personal_info.required` has entries that no longer exist in `analysisPlan.structuredDataSchema.properties.owner_personal_info.properties`" as noted in [schema discussions](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339019564838686740). ### Model-Related Issues There might be issues with the specific model being used. Some models have limitations with JSON schema validation or function calling. For example, there have been reports of "Client disconnected" errors with Anthropic models and JSON schema issues with Google models according to [model error reports](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1332510949940133960). ### OpenAI Changes Changes on OpenAI's end might be affecting your structured data output. In one case, a similar issue "was resolved when reverting to VAPI's default billing" as mentioned in [OpenAI-related issues](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339884806229458946). ### Token Limits
You might be exceeding token limits, which can prevent the generation of complete output. Most systems have maximum token limits to ensure performance and quality as noted in [token limitation discussions](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1332287510964994078). ### Formatting Issues The LLM might be outputting data in a format that's causing issues, such as using Markdown or duplicating URLs, which can interfere with proper structured data extraction according to [formatting issue reports](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339489458625708044). Would you like me to help you troubleshoot any of these specific areas, or do you have additional information about your setup that might help narrow down the cause? Sources: - [Discord Thread on Schema Issues](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1265637856383078461) - [Discord Thread on Schema Properties](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339019564838686740) - [Discord Thread on Model Errors](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1332510949940133960) - [Discord Thread on OpenAI Changes](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339884806229458946) - [Discord Thread on Token Limitations](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1332287510964994078) - [Discord Thread on Formatting Issues](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339489458625708044)
t
Same issue with this call id too: a336eba4-c9ea-4a37-ab27-3890db21d7fa
I just observed, in all these calls structured output completion tokens in 2000, seems like 2000 tokens is the max it generates. If this is true, is there possibility to increase that.
k
Thank you for reaching out to us. We've received your message and will get back to you as soon as possible.
Hi, checking if this is resolved/solved for you?
t
Hi, issue is not resolved
Could you please check
k
Can you try one more call by increasing the time of seconds to 30 so that the endorphic model gets enough time to generate the response?
Copy code
{
  "structuredDataPlan": {
    "enabled": true,
    "timeoutSeconds": 30, // Increase this value
    "schema": {
      // Your schema
    }
  }
}
t
Hey Shubham, Still the same error. I did put the timeoutSeconds as 60. Completion tokens stopped at 2000. Call id: 2af207b8-d7ed-4337-816f-9fdde5896300
k
🔵 04:40:48:136 Couldn't Parse JSON. Error: SyntaxError: Expected ',' or '\]' after array element in JSON at position 8507 (line 83 column 6). { "interview_completed": true, "reschedule_interview": { "reason": "", "day": 0, "mo
2 Views