Inconsistent / Hallucinated Responses During Usage

Hi team,

I wanted to report an issue I experienced while using the tool during a real-world scenario.

At one point, the responses started drifting from the original context and became inaccurate. The tool began generating answers that didn’t align with the prompt and felt inconsistent with what was expected.

This made it difficult to rely on the output, especially in a situation where accuracy really matters.

Expected behavior:
Accurate, context-aware, and consistent responses aligned with the prompt.

Actual behavior:
Responses became irrelevant / hallucinated and broke context mid-session.

Happy to provide more details if needed.