[adityasingh2400]

Please ensure you have read the contribution guide before creating a pull request.

Link to Issue or Description of Change

1. Link to an existing issue (if applicable):

• Closes: Interactions API requests chain onto rewound turns after Runner.rewind_async #7820

Problem:
On the Interactions API, the model's context is the server-side history behind previous_interaction_id. _find_previous_interaction_state chose that id by scanning raw session.events. After Runner.rewind_async, the next request therefore continued from a rewound turn's interaction, and the turn came back into the model's context. rewind.md promises that ADK "leaves the rewound turns out of every prompt that it builds afterwards".

Solution:
Scan _apply_rewinds(events) instead of events. _rewind_events describes _apply_rewinds as the single source of truth for which events are live, and prompt contents, compaction, memory ingestion and the agent router already use it. After a rewind, the request now continues from the last live interaction, or sends no previous_interaction_id when every interaction was rewound. _managed_agent.py uses the same helper, so it gets the fix as well. The branch filtering and the returned (interaction_id, environment_id) shape are unchanged.

Testing Plan

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

Added to tests/unittests/flows/llm_flows/context/test_interactions.py:

• test_find_previous_interaction_state_skips_rewound_turns: rewinding turn 2 makes the lookup return turn 1's (interaction_id, environment_id).
• test_find_previous_interaction_state_none_when_every_turn_rewound: returns (None, None).
• test_find_previous_interaction_state_chains_turns_after_a_rewind: a turn taken after the rewind is still found, as a guard.

# On main (before the fix)
AssertionError: assert ('int_2', 'env_2') == ('int_1', 'env_1')
AssertionError: assert ('int_1', None) == (None, None)
2 failed, 1 passed

# With the fix
$ pytest tests/unittests/flows/llm_flows/context/test_interactions.py
24 passed

$ pytest tests/unittests/flows tests/unittests/runners tests/unittests/agents
2216 passed, 1 skipped, 4 xfailed

pre-commit run passes on the changed files, and mypy reports no issues in _interactions.py.

Manual End-to-End (E2E) Tests:

I ran a real Runner with LlmAgent(model=Gemini("gemini-2.5-flash", use_interactions_api=True)) and an InMemorySessionService. Only interactions_utils.generate_content_via_interactions was stubbed: it records llm_request.previous_interaction_id and returns interaction_id=f"int_{n}". I ran turn 1 ("My name is Ada."), turn 2 ("Actually, call me Bob."), then rewind_async(rewind_before_invocation_id=<turn 2>), then turn 3 ("What is my name?"). The script is in #7820.

=== BEFORE (main) ===
previous_interaction_id per request: [None, 'int_1', 'int_2']
turn 3 continues from: int_2
=== AFTER (this PR) ===
previous_interaction_id per request: [None, 'int_1', 'int_1']
turn 3 continues from: int_1

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.

Additional context

None.