MAINT: Remove unused fields in MessagePiece

[behnam-o]

There are several fields in the message piece class that are never set or used. This PR removes them to make the constructor and objects cleaner.

There is a "scores" field, that gets populated during memory queries, and this behavior is kept as is, but replaced with a property with an internal setter.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Removes several unused MessagePiece fields and associated memory/query surface area, and updates tests/docs to match the simplified model.

Changes:

• Removed originator, scorer_identifier, and targeted_harm_categories from MessagePiece serialization and memory storage.
• Changed scores hydration to occur post-construction via MessagePiece._set_scores() (memory layer + tests updated).
• Removed targeted_harm_categories filtering capability from MemoryInterface.get_attack_results() and DB-specific implementations.

Reviewed changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
pyrit/models/message_piece.py	Removes unused fields; introduces score hydration via _set_scores; updates to_dict() output.
pyrit/memory/memory_models.py	Drops targeted_harm_categories persistence and sets scores via _set_scores() when hydrating MessagePiece.
pyrit/memory/memory_interface.py	Removes targeted_harm_categories from get_attack_results() and deletes abstract hook for harm-category query conditions.
pyrit/memory/sqlite_memory.py	Removes SQLite harm-category attack-result filtering helper.
pyrit/memory/azure_sql_memory.py	Removes Azure SQL harm-category attack-result filtering helper.
pyrit/score/conversation_scorer.py	Stops copying removed originator field when creating derived pieces.
tests/unit/models/test_message_piece.py	Updates tests for removed fields and new score-hydration pathway; deletes scorer/harm-category tests.
tests/unit/memory/memory_interface/test_interface_attack_results.py	Updates helper signature and removes harm-category filtering tests.
tests/unit/executor/attack/component/test_conversation_manager.py	Updates tests to attach scores via _set_scores() instead of constructor.
tests/integration/memory/test_azure_sql_memory_integration.py	Removes integration test for harm-category filtering.
tests/unit/prompt_normalizer/test_prompt_converter_configuration.py	Changes how mock converter “name” is set.
doc/code/memory/3_memory_data_types.md	Removes docs for some fields (but not all—see comments).
doc/code/memory/1_sqlite_memory.ipynb / doc/code/memory/9_exporting_data.ipynb / doc/code/executor/attack/1_prompt_sending_attack.ipynb	Updates notebook outputs to reflect new schema/fields and new execution environment.

[behnam-o]

will merge after #1625 MAINT: Emit deprecation warning for unused fields being removed from message piece

[hannahwestra25]

why do we get rid of scorer_identifier but not the other identifiers ?

[romanlutz]

Because scores should include it

[behnam-o]

This particular one, along with other fields here, is just never populated in other classes - and this PR is only focusing on such unused fields that are never populated, so there's no point in having them.

Other ones could be in scope for removal, for example attack_identifier is also duplicated between message pieces and attack_results ... but the difference is, that is populated in other classes today, and will have a more cascading change all over the repo.