Configuring Azure OpenAI with CrewAI: Correct way of setting up the embedder section of the TXTSearchTool

[palaemezie]

A few months ago, I was having difficulty finding the correct key that the embeddchain uses for the azure-openai provider using the CrewAI framework. So, I made a post about my challenge on the Azure AI Foundry Discord server; you can access the post via https://discordapp.com/channels/1113626258182504448/1113626258698412049/1366812163217428490. @amynic encouraged me to open a discussion (on the challenge I was facing) here. Although I have resolved the issue I faced, I would share the solution that worked for me here with the community.

[palaemezie]

After reviewing the CrewAI resource https://blog.crewai.com/configuring-azure-openai-with-crewai-a-comprehensive-guide/, I couldn't find where the correct key for the embeddchain was specified. So, I worked up this solution:

python

os.environ["OPENAI_API_KEY"] = os.environ.get("AZURE_OPENAI_API_KEY")
os.environ["AZURE_API_KEY"] = os.environ.get("AZURE_API_KEY")
os.environ["AZURE_API_BASE"] = os.environ.get("AZURE_API_BASE")

azure_config = dict(
        llm=dict(
            provider="azure_openai",
            config=dict(
                model="gpt-4o",
                deployment_name="gpt-4o",
                api_key=os.environ["AZURE_API_KEY"],
                base_url=os.environ["AZURE_API_BASE"],
                api_version="2025-01-01-preview",
                temperature=0.2,
            ),
        ),
        embedder=dict(
            provider="azure_openai",
            config=dict(
                model="text-embedding-ada-002",
                deployment_name="text-embedding-ada-002",
                api_key=os.environ["AZURE_API_KEY"],
                api_base=os.environ["AZURE_API_BASE"],
            )
        )
    )

search_tool = TXTSearchTool(
    config=azure_config,
    txt='ai.txt',
    overwrite=True,
    verbose=True
)

I deployed the GPT-4o model in Azure AI Foundry and used the text-embedding-ada-002 as my embedding model to embed the text file ai.txt.