Flaky test: test_cli_clear_resets_conversation_chain #270

test_cli_clear_resets_conversation_chain is flaky because it depends on LLM behavior.

Observed Failure

The test tells the model a "secret code", clears the conversation, then asks for it back. The model sometimes fails to recall it, causing the test to fail.

I'm sorry, but I don't have access to any secret codes from our previous interactions or in my current environment.

Root Cause

LLM-dependent tests are inherently non-deterministic. The model may:

• Not store the information as expected
• Not recall it reliably after context clear
• Respond differently based on temperature/sampling

Options

1. Remove the test - LLM behavior is outside our control
2. Make it more resilient - Retry logic, or check for partial matches
3. Mark as #[ignore] - Skip in regular CI, run manually
4. Use mock responses - Replace LLM call with deterministic mock for this test

Evidence

Failed in CI on PR #269 (unrelated glob test fix), passed on re-run.