autogen-assistant doesn't work on Azure OpenAI

[ArneJanning]

If you edit the the get_default_agent_config()-Method in utils.py to use Azure OpenAI, as described in the documentation:

`def get_default_agent_config(work_dir: str, skills_suffix: str = "") -> FlowConfig: """ Get a default agent flow config . """

llm_config = LLMConfig(
    config_list=[{
            "model": "gpt4",
            "api_key": "<myKey>",
            "api_base": "<myBase>",
            "api_type": "azure",
            "api_version": "2023-06-01-preview" 
    }],
    temperature=0,
)

` Then you'll receive an error:

openai.OpenAIError: 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

[victordibia]

Thanks for noting this @ArneJanning .

I am looking at the OAI_CONFIG_LIST documentation here and it looks like the correct formatting for azure openai models is

{
        "model": "<your Azure OpenAI deployment name>",
        "api_key": "<your Azure OpenAI API key here>",
        "base_url": "<your Azure OpenAI API base here>",
        "api_type": "azure",
        "api_version": "2023-07-01-preview"
    }

Note that the difference is api_base -> base_url.

What do you get when you try the above?

[ArneJanning]

Thank you for your answer @victordibia .

I changed the api_base to base_url and still get the same error:

`autogenra ui --port 8081

INFO: Started server process [105794]

INFO: Waiting for application startup.

INFO: Application startup complete.

INFO: Uvicorn running on http://127.0.0.1:8081 (Press CTRL+C to quit)

INFO: 127.0.0.1:56470 - "GET / HTTP/1.1" 200 OK

INFO: 127.0.0.1:56470 - "GET /app-a69e64a20490df34caf8.js HTTP/1.1" 200 OK

INFO: 127.0.0.1:56482 - "GET /framework-34c4775408c4a2a587eb.js HTTP/1.1" 200 OK

INFO: 127.0.0.1:56470 - "GET /webpack-runtime-9a1c1a0dc45832ebafbd.js HTTP/1.1" 200 OK

INFO: 127.0.0.1:56482 - "GET /page-data/index/page-data.json HTTP/1.1" 200 OK

INFO: 127.0.0.1:56470 - "GET /page-data/app-data.json HTTP/1.1" 200 OK

INFO: 127.0.0.1:56470 - "GET /component---src-pages-index-tsx-9c8c5671443f8293283f.js HTTP/1.1" 200 OK

INFO: 127.0.0.1:56470 - "GET /skills?user_id=guestuser@gmail.com HTTP/1.1" 200 OK

INFO: 127.0.0.1:56482 - "GET /messages?user_id=guestuser@gmail.com HTTP/1.1" 200 OK

INFO: 127.0.0.1:56488 - "GET /manifest.webmanifest HTTP/1.1" 200 OK

Traceback (most recent call last):

File "/home/arne/src/autogen3/autogen/samples/apps/autogen-assistant/autogenra/web/app.py", line 85, in add_message response_message: Message = chatmanager.chat( ^^^^^^^^^^^^^^^^^

File "/home/arne/src/autogen3/autogen/samples/apps/autogen-assistant/autogenra/autogenchat.py", line 24, in chat flow = AutoGenFlow(config=flow_config, history=history, work_dir=scratch_dir, asst_prompt=skills_suffix) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/arne/src/autogen3/autogen/samples/apps/autogen-assistant/autogenra/autogenflow.py", line 26, in init self.sender = self.load(config.sender) ^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/arne/src/autogen3/autogen/samples/apps/autogen-assistant/autogenra/autogenflow.py", line 116, in load agent = autogen.UserProxyAgent(**asdict(agent_spec.config)) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/arne/src/autogen3/.venv/lib/python3.11/site-packages/autogen/agentchat/user_proxy_agent.py", line 72, in init super().init(

File "/home/arne/src/autogen3/.venv/lib/python3.11/site-packages/autogen/agentchat/conversable_agent.py", line 115, in init self.client = OpenAIWrapper(**self.llm_config) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/arne/src/autogen3/.venv/lib/python3.11/site-packages/autogen/oai/client.py", line 77, in init self._clients = [self._client(config, openai_config) for config in config_list] # could modify the config ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/arne/src/autogen3/.venv/lib/python3.11/site-packages/autogen/oai/client.py", line 77, in self._clients = [self._client(config, openai_config) for config in config_list] # could modify the config ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/arne/src/autogen3/.venv/lib/python3.11/site-packages/autogen/oai/client.py", line 138, in _client client = OpenAI(**openai_config) ^^^^^^^^^^^^^^^^^^^^^^^

File "/home/arne/src/autogen3/.venv/lib/python3.11/site-packages/openai/_client.py", line 93, in init raise OpenAIError(

openai.OpenAIError: 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 `

[miiiz]

I get the same error with OpenAI api_key. No matter how I set get_default_agent_config(). I tried llm_config = LLMConfig( config_list=[{ "model": "gpt-4", "api_key": "<myKey>", }], temperature=0, ) and also tried the old way I use in autogen config_list = [ { 'model': 'gpt-4', 'api_key': "<myKey>", }, ] llm_config = { "cache_seed": 42, "temperature": 0, "config_list": config_list, }

[olivMertens]

Yeah i've the same issue, whatever i used, the message stay here the api_key is not passed to the original openai library It works with txt files with the config_list = autogen.config_list_openai_aoai(key_file_path=".\keys",exclude={"azure"}) If i haven't a open_ai key it didn't work :

chainlit = "^0.7.700" openai = "^1.3.6" langchain = "^0.0.343" pyautogen = "^0.2.0"

[miiiz]

I get the same error with OpenAI api_key. No matter how I set get_default_agent_config(). I tried llm_config = LLMConfig( config_list=[{ "model": "gpt-4", "api_key": "<myKey>", }], temperature=0, ) and also tried the old way I use in autogen config_list = [ { 'model': 'gpt-4', 'api_key': "<myKey>", }, ] llm_config = { "cache_seed": 42, "temperature": 0, "config_list": config_list, }

I notice that we must edit the config on the frontend and enter api_key there. That explains the keep coming error.

[ruanwz]

The api-version is missing, so I change lib/python3.11/site-packages/openai/_base_client.py at about line 863 to include this parameter. After this change, the azure model set in frontend works.

        url = request.url
        # set url parameter to api-version=2023-07-01-preview 
        url = url.copy_set_param('api-version', '2023-07-01-preview')

[codernew007]

I've tried coding like this, and then run python test/twoagent.py success.

config_list = [ { 'model': 'gpt-3.5-turbo', 'api_key': "xx","base_url":"xxx","api_type":"azure","api_version": "2023-07-01-preview" }]

[ArneJanning]

The api-version is missing, so I change lib/python3.11/site-packages/openai/_base_client.py at about line 863 to include this parameter. After this change, the azure model set in frontend works.

        url = request.url
        # set url parameter to api-version=2023-07-01-preview 
        url = url.copy_set_param('api-version', '2023-07-01-preview')

I'm sorry, I still don't get it. Where would you specify the Azure model in the frontend? And how? I tried to change getDefaultConfigFlows() in frontend/src/components/views/utils.ts to

export const getDefaultConfigFlows = () => { const llm_model_config: IModelConfig[] = [ { model: "gpt4-0613", api_key: "<key>", base_url: "<azure url>", api_type: "azure" }, ];

but without success.

The error openai.OpenAIError: 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 just doesn't go away, whatever I do on the frontend or the backend.

[ChunshengLin]

To fix this, try to change code at autogen-assistant/frontend/src/components/atoms.tsx. Add the following code, then cd autogen-assistant/frontend, use yarn build rebuild the frontend.

After that, when add mode, you need fill the Api Version, then save it.

Now you can use the Azure OpenAI.

[Li-Yanzhi]

When direct pip and run autogenra or autogenstudio from cmd, how to specify azure OpenAI config file? and I tried to edit the model config (add key/base_url, model type and version etc) from autogenstudio Build page, it did not take effect too. Maybe need to add a config parameter for this, e.g. autogenstudio ui --config "config_file"

[victordibia]

Looking into it. Thanks

Get Outlook for iOShttps://aka.ms/o0ukef

---

From: Yanzhi Li @.> Sent: Saturday, December 16, 2023 8:32:34 AM To: microsoft/autogen @.> Cc: Comment @.>; Assign @.> Subject: Re: [microsoft/autogen] autogen-assistant doesn't work on Azure OpenAI (Issue #755)

When direct pip and run autogenra or autogenstudio from cmd, how to specify azure OpenAI config file? and I tried to edit the model config (add key/base_url, model type and version etc) from autogenstudio Build page, it did not take effect too. Maybe need to add a config parameter for this, e.g. autogenstudio ui --config "config_file"

— Reply to this email directly, view it on GitHubhttps://github.com/microsoft/autogen/issues/755#issuecomment-1858861660 or unsubscribehttps://github.com/notifications/unsubscribe-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ou are receiving this email because you commented on the thread.

Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[dumbkid-tom]

my issue was llm_config was not passed to autogen.groupchatmanager function. After the fix, no more api_key missing error.

BEFORE:

groupchat = autogen.GroupChat(agents=[user_proxy, assistant], messages=[], max_round=MAX_ITER)

manager = autogen.GroupChatManager(groupchat=groupchat)

AFTER:

groupchat = autogen.GroupChat(agents=[user_proxy, assistant], messages=[], max_round=MAX_ITER)

manager = autogen.GroupChatManager(groupchat=groupchat, llm_config=llm_config)

Here is my llm_config:

config_list = [
    {
        "model": "gpt-35-turbo",  
        "api_key": os.environ.get("AZURE_OPENAI_API_KEY"),
        "base_url": os.environ.get("AZURE_OPENAI_API_BASE"),
        "api_type": "azure", 
        "api_version": "2023-07-01-preview", 
 }
]

llm_config={
    "timeout": 600,
    "cache_seed": 42,
    "config_list": config_list,
}

[huangpan2507]

When direct pip and run autogenra or autogenstudio from cmd, how to specify azure OpenAI config file? and I tried to edit the model config (add key/base_url, model type and version etc) from autogenstudio Build page, it did not take effect too. Maybe need to add a config parameter for this, e.g. autogenstudio ui --config "config_file"

I had the same doubt about using run autogenstudio ui from cmd

[victordibia]

HI @huangpan2507 ,

AutoGen studio is a UI interface on top of AutoGen. Currently, it is built such that the user specifies the agent workflow configuration in the UI (rather than a config file).