feat: add support for system messages to RunnableRails

[smruthi33]

Description

The current implementation does not support messages that have a System Message

guardrails_config = RailsConfig.from_path(<Path>)
guardrails = RunnableRails(guardrails_config)
client = ChatOpenAI(
                base_url=GEN_AI_ENDPOINT,
                api_key=GEN_AI_ENDPOINT_AUTHORIZATION,
                model= MODEL_NAME,
                temperature=0                
            )
messages = [
                (
                    "system", system
                )
                ,(
                    "human", prompt
                )
            ]
prompt = ChatPromptTemplate.from_messages(messages)

chain_with_guardrails =  prompt | (guardrails | client)

In this setup, RunnableRails does not handle messages containing system roles. Attempting to pass such a structure as a dictionary causes conflicts with the format expected by the Runnable interface from LangChain, which assumes a flat input structure and does not support multiple message types out of the box. This change addresses that limitation.

Related Issue(s)

Checklist

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

[Pouyanpi]

Thank you @smruthi33 for opening this PR 🚀

Would be great if you can add a test to test_runnable_rails.py. Thank you!

[Pouyanpi]

@tgasser-nv I added some tests. I think we are good to merge after your review:

I am not going to include following minimal integration test but it works.

import pytest
import os
from langchain_openai import ChatOpenAI
from langchain_core.messages import AIMessage, HumanMessage, SystemMessage
from langchain_core.prompt_values import ChatPromptValue

from nemoguardrails import RailsConfig
from nemoguardrails.integrations.langchain.runnable_rails import RunnableRails

requires_openai = pytest.mark.skipif(
    os.environ.get("OPENAI_API_KEY") is None,
    reason="Requires OPENAI_API_KEY environment variable",
)


@requires_openai
def test_system_message_with_openai():
    """Tests SystemMessage handling with a real OpenAI model."""
    try:
        llm = ChatOpenAI(model="gpt-4o", temperature=0)
    except ImportError:
        pytest.skip("langchain-openai not installed")
    except Exception as e:
        pytest.skip(f"Could not initialize ChatOpenAI: {e}")
        
    config = RailsConfig.from_content(config={"models": []})
    model_with_rails = RunnableRails(config)
    chain = model_with_rails | llm
    input_messages = [
        SystemMessage(
            content="You are an assistant that always replies starting with 'LOLOLOLO!'"
        ),
        HumanMessage(content="what can you do?"),
    ]
    result = chain.invoke(input=ChatPromptValue(messages=input_messages))
    assert isinstance(result, AIMessage)
    assert result.content.lower().startswith("lolololo!")

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 68.00%. Comparing base (67b8b7a) to head (c7707ab).

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #1106   +/-   ##
========================================
  Coverage    68.00%   68.00%           
========================================
  Files          161      161           
  Lines        15793    15795    +2     
========================================
+ Hits         10740    10742    +2     
  Misses        5053     5053           

Flag	Coverage Δ
python	68.00% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.