fix(llmrails): skip output rails when dialog disabled and no bot_message provided

[Pouyanpi]

When generate_async is called with options={"dialog": False, "output": True} and no bot_message is provided in context, output rails were incorrectly running and checking None. This fix ensures output rails only run when there's actual output to check.

Test matrix:

input	output	dialog	expect_input	expect_output
T	T	T	T	T
T	T	F	T	F ← bug fix
T	F	T	T	F
T	F	F	T	F
F	T	T	F	T
F	T	F	F	F ← bug fix
F	F	T	F	F
F	F	F	F	F

[greptile-apps]

Greptile Overview

Greptile Summary

Fixed a bug where output rails were incorrectly attempting to validate None when dialog=False and no bot_message was provided in the context. The fix adds a check for $bot_message is None to skip output rails in this scenario.

• Modified run dialog rails flow in llm_flows.co:32 to check both output == False OR $bot_message is None before skipping output rails
• Added comprehensive parametrized test with 8 combinations of input/output/dialog options to validate the fix
• The test matrix confirms output rails now correctly skip when dialog=False and there's no bot message to check

Confidence Score: 5/5

• This PR is safe to merge with minimal risk
• The fix is surgical and well-tested: it adds a single null check to prevent output rails from running on None values, which addresses the exact bug described. The comprehensive parametrized test validates all 8 combinations of rail options, ensuring no regression in existing behavior while fixing the reported issue.
• No files require special attention

Important Files Changed

File Analysis

Filename	Score	Overview
nemoguardrails/rails/llm/llm_flows.co	5/5	Fixed output rails logic to check for None $bot_message when dialog is disabled, preventing null checks
tests/test_generation_options.py	5/5	Added comprehensive parametrized test covering all 8 combinations of input/output/dialog rail options

Sequence Diagram

sequenceDiagram
    participant User
    participant generate_async
    participant run_dialog_rails
    participant process_bot_message
    participant output_rails

    User->>generate_async: "Hello!" with options{dialog: False, output: True}
    Note over generate_async: No bot_message in context<br/>($bot_message is None)
    generate_async->>run_dialog_rails: Process with generation_options
    
    alt dialog == False AND (output == False OR $bot_message is None)
        Note over run_dialog_rails: NEW: Check added for $bot_message is None
        run_dialog_rails->>User: Skip output rails<br/>Return StartUtteranceBotAction($user_message)
    else dialog == False AND output == True AND $bot_message exists
        run_dialog_rails->>process_bot_message: Create BotMessage($bot_message)
        process_bot_message->>output_rails: Run output rails on $bot_message
        output_rails->>User: Return processed message
    else dialog == True
        Note over run_dialog_rails: Normal flow: generate intent & bot message
        run_dialog_rails->>process_bot_message: Continue usual process
        process_bot_message->>output_rails: Run output rails
        output_rails->>User: Return processed message
    end

Loading

[greptile-apps]

_{2 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!