fix: Prevent explain_info overwrite during stream_async

[andompesta]

Description

This PR addresses an issue where self.explain_info could be overwritten during calls to stream_async, potentially leading to incomplete explanation data.

The root cause was that the generate_async task and the _run_output_rails_in_streaming function (if output streaming rails were active) could independently initialize ExplainInfo objects if the explain_info_var context variable was not set prior to the generate_async task's creation.

The fix ensures that explain_info_var and self.explain_info are initialized at the beginning of the stream_async method. This guarantees that both the generate_async task (via context variable inheritance) and _run_output_rails_in_streaming use the same ExplainInfo instance, preserving data integrity.

[same as https://github.com//pull/1193 but with pre-commit hooks done]

Related Issue(s)

#1192

Checklist

• [ x ] I've read the CONTRIBUTING guidelines.
• I've updated the documentation if applicable.
• [ x ] I've added tests if applicable.
• [ x ] @mentions of the person or team responsible for reviewing proposed changes.

[Copilot]

Pull Request Overview

This PR fixes the missing LLM call tracking in the explain info and refactors its initialization logic while adding tests to ensure proper behavior with pre-commit hooks.

• Added asynchronous tests to validate that the explain info correctly records LLM calls.
• Refactored llmrails.py to encapsulate explain info initialization in the _ensure_explain_info() method and removed redundant updating functions.

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
tests/test_llm_rails_context_variables.py	Introduces tests for LLM call tracking and verifies expected call counts.
nemoguardrails/rails/llm/llmrails.py	Refactors and unifies the logic for initializing explain info, removing redundant code.

Comments suppressed due to low confidence (1)

tests/test_llm_rails_context_variables.py:111

• The comment above this assertion states that 6 LLM calls are expected, but the assertion checks for 5. Please update either the comment or the expected value to ensure consistency.

assert (len(chat.app.explain().llm_calls) == 5), "number of llm call not as expected. Expected 5, found {}".format(len(chat.app.explain().llm_calls))

[Pouyanpi]

LGTM! Thank you @andompesta