The community member is experiencing an error while using the llamaindex subquestionqueryengine with the guidance question generator, despite following the demo shown on the llamaindex blog. Another community member suggests that the issue is likely due to the language model not producing a valid JSON object, and recommends using a tool that takes advantage of OpenAI's function/tool calling API, such as a Pydantic program or Pydantic query engine, instead of relying on JSON parsing. The community member with OpenAI access through Azure notes that the Azure platform does not yet have function calling capability, and another community member expresses dissatisfaction with the performance of the Guidance API.
While using llamaindex subquestionqueryengine with guidance question generator, i am getting this error. I am following the demo showed over the llamaindex blog exactly but getting this error everytime.
This is because the LLM decided not produce a valid JSON object -- it happens.
Tbh I would probably just use something that takes advantages of openais function/tool calling API rather than relying on JSON parsing. Something like a pydantic program or pydantic query engine
