'AzureOpenAI' object has no attribute 'create'

[VaibhavChemboli116]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

from langchain_experimental.agents import create_csv_agent
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(client=client)

agent = create_csv_agent(
    llm, 
    r'ROOT_v1/Data/data_sample.csv',
    agent_type="openai-tools",
    allow_dangerous_code=True,
    verbose=True
)
Error occurred here -> agent.run("how many rows are there?")

Description

I am trying to run the mentioned langchain agent. I have gpt-4o model and text-emdedding-ada-002 model deployed in my azure account. I am able to access both the models using AzureOpenAI and I am able to create embeddings and get response too. But if I try to build agent I am facing the mentioned error.

System Info

Name: langchain
Version: 0.2.12

Name: langchain-community
Version: 0.2.10

Name: langchain-experimental
Version: 0.0.64

Name: openai
Version: 1.40.3

[dosubot[bot]]

Hey there, @VaibhavChemboli116! I'm here to help you with the issue you're facing. I'm a bot designed to assist with bug fixes, answer questions, and guide you in becoming a contributor. Let's tackle this together!

The error occurs because the AzureOpenAI class does not have a create method. Instead, you should use the completions attribute of the AzureOpenAI object. Here's how you can modify your code:

1. Initialize the AzureOpenAI object correctly:

   from langchain_openai import AzureOpenAI
   
   llm = AzureOpenAI(
       deployment_name="gpt-4o",
       temperature=0,
       azure_api_key="your-api-key",
       azure_endpoint="https://your-endpoint.openai.azure.com/",
       openai_api_version="2023-05-15"
   )

2. Use the completions attribute to create a chat completion:

   response = llm.client.completions.create(
       model="gpt-4o",
       messages=[
           {"role": "system", "content": "You are a helpful assistant."},
           {"role": "user", "content": "What is the capital of France?"}
       ]
   )
   print(response)

3. Integrate with create_csv_agent:

   from langchain_experimental.agents import create_csv_agent
   from langchain_openai import AzureOpenAI
   
   llm = AzureOpenAI(
       deployment_name="gpt-4o",
       temperature=0,
       azure_api_key="your-api-key",
       azure_endpoint="https://your-endpoint.openai.azure.com/",
       openai_api_version="2023-05-15"
   )
   
   agent = create_csv_agent(
       llm, 
       r'ROOT_v1/Data/data_sample.csv',
       agent_type="openai-tools",
       allow_dangerous_code=True,
       verbose=True
   )
   
   agent.run("how many rows are there?")

Make sure to replace "your-api-key" and "https://your-endpoint.openai.azure.com/" with your actual Azure OpenAI API key and endpoint. This should resolve the error and allow you to run the agent successfully ^[1][2].

To continue talking to Dosu, mention @dosu.

[VaibhavChemboli116]

llm = AzureOpenAI(
api_version=api_version,
azure_endpoint=endpoint,
api_key=api_key,
azure_ad_token_provider=token_provider,
)
agent = create_csv_agent(
llm,
r'/mnt/batch/tasks/shared/LS_root/mounts/clusters/sashirekhar1/code/ROOT_v1/Data/data_sample.csv',
agent_type="openai-tools",
allow_dangerous_code=True,
verbose=True
)
AttributeError: 'AzureOpenAI' object has no attribute 'bind'

[dosubot[bot]]

To resolve the AttributeError: 'AzureOpenAI' object has no attribute 'bind' error, you should use the bind_tools method instead of bind. The AzureOpenAI class does not have a bind method, but it does have a bind_tools method for binding tools.

Here is an example of how you can modify your code:

1. Initialize the AzureOpenAI object correctly:

   from langchain_openai import AzureOpenAI
   
   llm = AzureOpenAI(
       api_version=api_version,
       azure_endpoint=endpoint,
       api_key=api_key, 
       azure_ad_token_provider=token_provider,
   )

2. Use the bind_tools method to bind tools:

   from langchain_experimental.agents import create_csv_agent
   
   agent = create_csv_agent(
       llm, 
       r'/mnt/batch/tasks/shared/LS_root/mounts/clusters/sashirekhar1/code/ROOT_v1/Data/data_sample.csv',
       agent_type="openai-tools",
       allow_dangerous_code=True,
       verbose=True
   )
   
   # Bind tools using bind_tools method
   runnable = llm.bind_tools(agent.tools)
   
   # Now you can use the runnable object
   response = runnable.invoke("how many rows are there?")
   print(response)

Make sure to replace the placeholders with your actual values. This should resolve the error and allow you to run the agent successfully ^[1][2][3].

To continue talking to Dosu, mention @dosu.

[VaibhavChemboli116]

llm = AzureOpenAI(
api_version=api_version,
azure_endpoint=endpoint,
api_key=api_key,
azure_ad_token_provider=token_provider,
)
runnable = llm.bind_tools(agent.tools)

agent = create_csv_agent(
llm,
r'/mnt/batch/tasks/shared/LS_root/mounts/clusters/sashirekhar1/code/ROOT_v1/Data/data_sample.csv',
agent_type="openai-tools",
allow_dangerous_code=True,
verbose=True
)

AttributeError Traceback (most recent call last)
Cell In[25], line 26
3 from langchain_openai import ChatOpenAI
5 #creating llm using all the three ways is causing errors
6
7 #llm = ChatOpenAI(client=client)
(...)
24
25 # Bind tools using bind_tools method
---> 26 runnable = llm.bind_tools(agent.tools)
28 agent = create_csv_agent(
29 llm,
30 r'/mnt/batch/tasks/shared/LS_root/mounts/clusters/sashirekhar1/code/ROOT_v1/Data/data_sample.csv',
(...)
33 verbose=True
34 )

AttributeError: 'AzureOpenAI' object has no attribute 'bind_tools'
I think we face issues when we create llm with AzureOpenAI, can you create llm with ChatOpenAI

[dosubot[bot]]

To avoid the AttributeError: 'AzureOpenAI' object has no attribute 'bind_tools' error, you should use ChatOpenAI instead of AzureOpenAI. Here is how you can modify your code:

1. Initialize the ChatOpenAI object:

   from langchain_openai import ChatOpenAI
   
   llm = ChatOpenAI(
       model_name="gpt-3.5-turbo",
       temperature=0,
       openai_api_key="your-api-key"
   )

2. Create the CSV agent:

   from langchain_experimental.agents import create_csv_agent
   
   agent = create_csv_agent(
       llm, 
       r'/mnt/batch/tasks/shared/LS_root/mounts/clusters/sashirekhar1/code/ROOT_v1/Data/data_sample.csv',
       agent_type="openai-tools",
       allow_dangerous_code=True,
       verbose=True
   )
   
   # Now you can use the agent
   response = agent.run("how many rows are there?")
   print(response)

Make sure to replace "your-api-key" with your actual OpenAI API key. This should resolve the error and allow you to run the agent successfully ^[1][2].

To continue talking to Dosu, mention @dosu.

[VaibhavChemboli116]

from langchain_openai import ChatOpenAI
from azure.identity import DefaultAzureCredential, get_bearer_token_provider

token_provider = get_bearer_token_provider(DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default")

llm = ChatOpenAI(
model_name="gpt-4o-2024-07-23 ",
openai_api_key=os.getenv("OPENAI_KEY"),
#azure_ad_token_provider=token_provider
)
response = llm.chat.completions.create(
model="gpt-4o",
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "What is the capital of France?"}
]
)
print(response)

AttributeError Traceback (most recent call last)
Cell In[38], line 1
----> 1 response = llm.chat.completions.create(
2 model="gpt-4o",
3 messages=[
4 {"role": "system", "content": "You are a helpful assistant."},
5 {"role": "user", "content": "What is the capital of France?"}
6 ]
7 )
8 print(response)

AttributeError: 'ChatOpenAI' object has no attribute 'chat'
I think there isn't any option available to create langchain agent with azure