[BUG]

[annamalaiarunachalam]

Description

even if I set up the agent as Gemini, and pass the relevant API keys, in some cases, the crew when its run looks for openAI. why did you guys default everything to OpenAI, why can't the LLM be configured by the user for every instance where it is required??????

File ~\AppData\Roaming\Python\Python312\site-packages\litellm\llms\openai\openai.py:368, in OpenAIChatCompletion._get_openai_client(self, is_async, api_key, api_base, api_version, timeout, max_retries, organization, client)
366 else:
...
363 )
364 elif "Mistral API raised a streaming error" in error_str:
365 exception_mapping_worked = True

AuthenticationError: litellm.AuthenticationError: AuthenticationError: OpenAIException - The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable
Output is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...

Steps to Reproduce

no steps, just running crewAI's tutorial code - but using gemini models, of course with gemini API keys. I DO NOT USE OPENAI.

Expected behavior

accept other models.

Screenshots/Code snippets

none

Operating System

Ubuntu 20.04

Python Version

3.10

crewAI Version

0.108.0

crewAI Tools Version

0.38.1

Virtual Environment

Venv

Evidence

File ~\AppData\Roaming\Python\Python312\site-packages\litellm\llms\openai\openai.py:368, in OpenAIChatCompletion._get_openai_client(self, is_async, api_key, api_base, api_version, timeout, max_retries, organization, client)
366 else:
...
363 )
364 elif "Mistral API raised a streaming error" in error_str:
365 exception_mapping_worked = True

AuthenticationError: litellm.AuthenticationError: AuthenticationError: OpenAIException - The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable
Output is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...

Possible Solution

support all models, remove openai as default for everything

Additional context

none

[Vidit-Ostwal]

Can you share you entire code?
@annamalaiarunachalam

[annamalaiarunachalam]

it's the deeplearning.ai code and today I tried from your docs >> a serperapi sample. same issue. You will be able to understand what I mean. I had to assign llm, manager_llm, planning_llm or any such properties of a given agent. the agent by default uses openAI, My point is instead of defaulting OpenAI for everything, you should allow us to setup a model as for the entire application or a set of models. #create agents researcher = Agent( role='Generative AI Research Analyst', goal='Provide up-to-date information about the latest innovations from the AI industry', backstory='An expert analyst with a keen eye for technology trends.', tools=[search_tool], verbose=True, ) # Define tasks research = Task( description='Research the latest trends in the AI industry and provide a summary.', expected_output='A summary of the top 3 trending developments in the AI industry with a unique perspective on their significance.', agent=researcher ) # Assemble a crew with planning enabled crew = Crew( agents=[researcher], tasks=[research], verbose=True, planning=True, # Enable planning feature ) # Execute tasks crew.kickoff()

…

On Thu, 3 Apr 2025 at 22:59, Vidit Ostwal ***@***.***> wrote: Can you share you entire code? @annamalaiarunachalam <https://github.com/annamalaiarunachalam> — Reply to this email directly, view it on GitHub <#2517 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALS6KGGIKNVOWJSHCG3I6WL2XVEERAVCNFSM6AAAAAB2L6MIPOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZWGA3TONJUGY> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: Vidit-Ostwal]*Vidit-Ostwal* left a comment (crewAIInc/crewAI#2517) <#2517 (comment)> Can you share you entire code? @annamalaiarunachalam <https://github.com/annamalaiarunachalam> — Reply to this email directly, view it on GitHub <#2517 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALS6KGGIKNVOWJSHCG3I6WL2XVEERAVCNFSM6AAAAAB2L6MIPOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZWGA3TONJUGY> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[Vidit-Ostwal]

You can define your own LLM, and send it with llm parameter in the crew

Check out this doc:
https://docs.crewai.com/concepts/llms#google

Let me know if this works

from crewai import LLM

llm = LLM(
    model="gemini/gemini-1.5-pro-latest",
    temperature=0.7,
    vertex_credentials=vertex_credentials_json
)

[annamalaiarunachalam]

ofcourse, Iknow I can pass llm=<whatever llm I want> but it is not required only in the agent, it looks like I have to provide that in crew as well, there are several properties that require a model. I have to figure out where the error occurs and then provide the llm detail for each of those. you still refuse to provide a reason for choosing OpenAI as a default LLM - I wonder why?

…

On Fri, 4 Apr 2025 at 19:24, Vidit Ostwal ***@***.***> wrote: You can define your own LLM, and send it with llm parameter in the crew Check out this doc: https://docs.crewai.com/concepts/llms#google Let me know if this works — Reply to this email directly, view it on GitHub <#2517 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALS6KGDPB527IL3IC5KJ6FL2XZTWFAVCNFSM6AAAAAB2L6MIPOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZYGQYTAMJYGU> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: Vidit-Ostwal]*Vidit-Ostwal* left a comment (crewAIInc/crewAI#2517) <#2517 (comment)> You can define your own LLM, and send it with llm parameter in the crew Check out this doc: https://docs.crewai.com/concepts/llms#google Let me know if this works — Reply to this email directly, view it on GitHub <#2517 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALS6KGDPB527IL3IC5KJ6FL2XZTWFAVCNFSM6AAAAAB2L6MIPOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZYGQYTAMJYGU> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[Vidit-Ostwal]

Yes, you gotta initialise each one of them.
TIP: You can just initialize one of them and assign other the same llm. This should work.

you still refuse to provide a reason for choosing OpenAI as a default LLM -
I wonder why?

OpenAI’s API became the default because it was one of the first and most widely adopted. Other companies, like DeepSeek, have made their APIs compatible with it for easier integration. Think of it like a fallback in function parameters — if nothing is specified, it defaults to the most established option.

This is my personal opinion, I think this could be much better answered by maintainers.

[annamalaiarunachalam]

for instance, I get error for the following: from crewai_tools import ( FileReadTool, ScrapeWebsiteTool, MDXSearchTool, SerperDevTool ) search_tool = SerperDevTool() scrape_tool = ScrapeWebsiteTool() read_resume = FileReadTool(file_path='./anna_cv.md') semantic_search_resume = MDXSearchTool(mdx='./anna_revised.md', llm=llm_goog1) There is not attribute called LLM in the MDXSearchTool class, how do I add. it defaults to OpenAI. --------------------------------------------------------------------------- KeyError Traceback (most recent call last) Cell In[7], line 11 9 scrape_tool = ScrapeWebsiteTool() 10 read_resume = FileReadTool(file_path=' ./anna_cv.md') ---> 11 semantic_search_resume = MDXSearchTool(mdx=' ./anna_revised.md', 12 llm=llm_goog1) File ~\AppData\Roaming\Python\Python312\site-packages\crewai_tools\tools\mdx_seach_tool\mdx_search_tool.py:32, in MDXSearchTool.__init__(self, mdx, **kwargs) 31 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/mdx_seach_tool/mdx_search_tool.py:31> def __init__(self, mdx: Optional[str] = None, **kwargs): ---> 32 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/mdx_seach_tool/mdx_search_tool.py:32> super().__init__(**kwargs) 33 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/mdx_seach_tool/mdx_search_tool.py:33> if mdx is not None: 34 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/mdx_seach_tool/mdx_search_tool.py:34> kwargs["data_type"] = DataType.MDX [... skipping hidden 1 frame] File ~\AppData\Roaming\Python\Python312\site-packages\crewai_tools\tools\rag\rag_tool.py:46, in RagTool._set_default_adapter(self) 42 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/rag/rag_tool.py:42> from embedchain import App 44 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/rag/rag_tool.py:44> from crewai_tools.adapters.embedchain_adapter import EmbedchainAdapter ---> 46 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/rag/rag_tool.py:46> app = App.from_config(config=self.config) if self.config else App() 47 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/rag/rag_tool.py:47> self.adapter = EmbedchainAdapter( 48 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/rag/rag_tool.py:48> embedchain_app=app, summarize=self.summarize 49 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/rag/rag_tool.py:49> ) 51 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/crewai_tools/tools/rag/rag_tool.py:51> return self ... 21 <https://file+.vscode-resource.vscode-cdn.net/c%3A/Users/AnnamalaiArunachalam/PythonProjects/CrewAI/~/AppData/Roaming/Python/Python312/site-packages/embedchain/embedder/openai.py:21> if api_key is None and os.getenv("OPENAI_ORGANIZATION") is None: File <frozen os>:714, in __getitem__(self, key) KeyError: 'OPENAI_API_KEY' If it was possible to set the model that I want to work with as a global var, and if that variable was accessible to all the agents, crews and whatever class that needs, it would have been simple. anyway, let me know how to handle the above error. thanks

…

On Fri, 4 Apr 2025 at 23:21, Vidit Ostwal ***@***.***> wrote: Yes, you gotta initialise each one of them. TIP: You can just initialize one of them and assign other the same llm. This should work. you still refuse to provide a reason for choosing OpenAI as a default LLM - I wonder why? OpenAI’s API became the default because it was one of the first and most widely adopted. Other companies, like DeepSeek, have made their APIs compatible with it for easier integration. Think of it like a fallback in function parameters — if nothing is specified, it defaults to the most established option. This is my personal opinion, I think this could be much better answered by maintainers. — Reply to this email directly, view it on GitHub <#2517 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALS6KGFUVX46ZPZXKAP27PD2X2PQ3AVCNFSM6AAAAAB2L6MIPOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZZGA2TKMJQGA> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: Vidit-Ostwal]*Vidit-Ostwal* left a comment (crewAIInc/crewAI#2517) <#2517 (comment)> Yes, you gotta initialise each one of them. TIP: You can just initialize one of them and assign other the same llm. This should work. you still refuse to provide a reason for choosing OpenAI as a default LLM - I wonder why? OpenAI’s API became the default because it was one of the first and most widely adopted. Other companies, like DeepSeek, have made their APIs compatible with it for easier integration. Think of it like a fallback in function parameters — if nothing is specified, it defaults to the most established option. This is my personal opinion, I think this could be much better answered by maintainers. — Reply to this email directly, view it on GitHub <#2517 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALS6KGFUVX46ZPZXKAP27PD2X2PQ3AVCNFSM6AAAAAB2L6MIPOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZZGA2TKMJQGA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[Vidit-Ostwal]

you need to pass a config paramete while initialising,

tool = MDXSearchTool(
    config=dict(
        llm=dict(
            provider="ollama", # Options include google, openai, anthropic, llama2, etc.
            config=dict(
                model="llama2",
                # Optional parameters can be included here.
                # temperature=0.5,
                # top_p=1,
                # stream=true,
            ),
        ),
        embedder=dict(
            provider="google", # or openai, ollama, ...
            config=dict(
                model="models/embedding-001",
                task_type="retrieval_document",
                # Optional title for the embeddings can be added here.
                # title="Embeddings",
            ),
        ),
    )
)

check out this doc page: https://docs.crewai.com/tools/mdxsearchtool#customization-of-model-and-embeddings

[PedroPini]

@annamalaiarunachalam The easiest fix — without needing to change any code — is to configure it the way you did and set your Mistral API key as OPENAI_API_KEY.
The reason is that embedchain defaults to using OpenAIEmbedder, so it expects that variable.

Here’s a working example using openrouter which is same approach for any provider:

OPENAI_API_KEY=same-key
OPENROUTER_API_KEY=same-key
OPENROUTER_API_BASE=https://openrouter.ai/api/v1
OPENROUTER_MODEL_NAME=openrouter/deepseek/deepseek-r1