AI Chatbot Bug Fixer Needed

Our AI chatbot has started acting strangely. Sometimes it gives the right answer. Other times it ignores information it already has, mixes things up, or answers with things that don't seem to come from our data. We're not sure where the problem is. It could be the frontend, backend, prompts, context handling, retrieval, application logic, or something else in the request flow. We're looking for someone who can investigate the whole workflow, find the real cause, and fix it if it's within the scope of this job. We're not looking for random prompt changes—we want someone who's good at tracking down bugs. ### What you'll do * Review the chatbot workflow from frontend to backend. * Find where the incorrect behavior is being introduced. * Explain the root cause. * Implement a fix if it's a straightforward change. * Leave a short summary of what was changed. ### Our stack * Python * OpenAI API * JSON * AI chatbot * Frontend + backend application If we uncover additional work, we'd be happy to continue the engagement. When applying, please share a similar chatbot or LLM issue you've diagnosed, how you identified the root cause, and how you fixed it.