[Pushpak Siva Sai (Pushpak731)]

Closes #3385

Problem

LLMGenericTextConverter forwards **kwargs (_prompt_kwargs) to both templates when rendering, but _build_identifier() hashed only the unrendered templates. Two converters built from the same template with different parameters — the docstring's own language=... example — shared one identifier: the second ConverterRegistry registration raised ValueError: Instance 'LLMGenericTextConverter::<hash>' already exists, and memory/analytics could not tell the configurations apart. Subclasses (ToneConverter, TranslationConverter, …) already promote their own parameter, so this hit the documented direct-use path.

Fix

The resolved template parameters are included in the identifier under a nested prompt_template_kwargs key:

• nested under one key so a template parameter can never collide with system_prompt_template_hash / user_prompt_template_hash;
• self._prompt_kwargs or None collapses the empty case to None, mirroring the extra_parameters fix for AzureMLChatTarget (BUG: AzureMLChatTarget identifier ignores the extra model parameters it sends #3331 / FIX Include the extra model parameters in the AzureMLChatTarget identifier #3332);
• subclasses' promoted parameters stay alongside it.

Tests (8 new)

• different kwargs → different unique_name and hash (the issue's scenario)
• identical kwargs → stable hash
• params expose prompt_template_kwargs == {"language": "French"}
• no kwargs → params.get("prompt_template_kwargs") is None
• user-prompt-template converters distinguished by kwargs too
• structured kwargs (lists) distinguish and stay JSON-serializable
• the issue's registry failure mode: two registrations succeed
• subclass non-regression: ToneConverter keeps its promoted tone param and distinct hashes

Verification

• tests/unit/converter/ — 2776 passed, 74 skipped
• tests/unit/registry/ + tests/unit/models/identifiers/ + identifier filters — 1114 passed
• ruff check + format clean

I used AI to assist me with this change.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.