I'm currently integrating Azure OpenAI Assistant into a project and facing difficulties with parsing uploaded files in the playground. Despite trying multiple formats such as .txt, .pdf, and .md, I consistently encounter parsing errors when the Assistant tries to read the files, both when running my code and directly through the Azure playground UI.
Error Description:
After uploading the files and initiating the process, the playground returns a full-stack trace indicating a parsing failure. Subsequently, the Assistant provides a response that seems to be a mix of hallucinations, containing little to no context related to the uploaded document.
Troubleshooting Steps:
- Checked the integrity and format of the uploaded documents.
- Attempted the process with different file formats.
- Ensured the documents are text-based and not image-based PDFs.
- Ran the same setup using OpenAI's standard API where it executed as expected.
Observations:
- The files appear to upload correctly.
- The error triggers a stack trace in the playground.
- Similar setups work correctly in OpenAI's non-Azure environment.
Additional Context:
- Azure OpenAI Assistant is in a beta release.
- I am attempting to process the files for a simple document retrieval and Q&A task. Below is an example of the stack trace received (error details and traceback):
