Fix langchain-google-genai integration by stripping 'models/' prefix

[devin-ai-integration]

Fix langchain-google-genai integration by stripping 'models/' prefix

Summary

Fixes #3702 - langchain-google-genai integration failing with "LLM Provider NOT provided" error.

The issue occurs when using langchain-google-genai's ChatGoogleGenerativeAI with CrewAI. The langchain library stores the model name with a models/ prefix (e.g., models/gemini/gemini-pro) which is used internally by Google's API, but LiteLLM doesn't recognize this format and throws an error.

Changes:

• Modified create_llm() in llm_utils.py to strip the models/ prefix when extracting model names from langchain-like objects
• Added 4 comprehensive unit tests covering various scenarios (with prefix, without prefix, case sensitivity)
• All existing tests continue to pass

The fix is backward compatible - models without the prefix are unchanged.

Review & Testing Checklist for Human

⚠️ CRITICAL - This PR is based on assumptions and needs real-world validation:

• Most Important: Install langchain-google-genai and test the actual failing case from issue langchain-google-genai integration fails with LLM Provider NOT provided on macOS #3702. Create an agent with ChatGoogleGenerativeAI(model="gemini/gemini-pro") and verify it works now.
• Test with other langchain integrations (e.g., langchain-openai) to ensure no regressions
• Verify the case-sensitivity logic is correct (should we strip uppercase Models/ too?)
• Check if other similar prefixes need to be handled (I only implemented models/ based on the issue)
• Run a full crew execution with langchain-google-genai end-to-end to ensure it works in practice

Test Plan Recommendation

from crewai import Agent, Task, Crew
from langchain_google_genai import ChatGoogleGenerativeAI
import os

# Set your GOOGLE_API_KEY
os.environ["GOOGLE_API_KEY"] = "your-key"

llm = ChatGoogleGenerativeAI(model="gemini/gemini-pro")

agent = Agent(
    role="Test Agent",
    goal="Test the fix",
    backstory="Testing langchain-google-genai integration",
    llm=llm
)

task = Task(
    description="Say hello",
    expected_output="A greeting",
    agent=agent
)

crew = Crew(agents=[agent], tasks=[task])
result = crew.kickoff()
print(result)

Notes

• Limitation: I wasn't able to test with the actual langchain-google-genai package due to environment constraints. The fix is based on analyzing the error message and creating mock objects that simulate the expected behavior.
• The fix only strips lowercase models/ prefix (case-sensitive) - this seemed like the safe default but should be validated.
• Unit tests all pass, but integration testing with the real library is essential before merging.

---

Link to Devin run: https://app.devin.ai/sessions/f38edeb68fa7484faaa5c6f9d5329afe
Requested by: João ([email protected])

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring