Need help getting started

Google AI Studio
,
1

I’m just getting started here.
I generated the key form AI Studio and typed in a question at the prompt (the Monty Hall problem). Got an answer. Then clicked on “Get SDK Code”. This gave me Python code. So far, so good.
But, the generated code is:

def generate():
    client = genai.Client(
        api_key=os.environ.get("GEMINI_API_KEY"),
    )

    model = "gemini-2.0-flash-lite"
    contents = [
        types.Content(
            role="user",
            parts=[
                types.Part.from_text(text="""There is a game show where there are three closed doors. <snip>
"""),
            ],
        ),
        types.Content(
            role="model",
            parts=[
                types.Part.from_text(text="""The best choice is to **swap doors**. Here's why: <snip>
"""),
            ],
        ),
        types.Content(
            role="user",
            parts=[
                types.Part.from_text(text="""INSERT_INPUT_HERE"""),
            ],
        ),
    ]
    generate_content_config = types.GenerateContentConfig(
    )

    for chunk in client.models.generate_content_stream(
        model=model,
        contents=contents,
        config=generate_content_config,
    ):
        print(chunk.text, end="")

The second types.Content element contains the answer. Why? Isn’t there supposed to be a requests.get(url) call to fetch the answer using a REST call? Is the client.models.generate_content_stream() call meant for the actual fetch?
What is the third types.Content doing? Seems to be empty.

3

Hi @Hussain_Akbar Welcome to the community
The code you shared uses the Gemini client library to interact with the API, so it doesn’t need requests.get().
The client.models.generate_content_stream() function handles the communication directly.
The code also includes your question and the model’s answer as a conversation history using two types.Content elements. The third types.Content element is a placeholder, allowing you to easily add a new question to continue the conversation.
Check this out GitHub - google-gemini/cookbook: Examples and guides for using the Gemini API.

4

@Pannaga_J Thanks for the reply. So, for a fresh query from scratch, I should remove the first two types.Contents elements and put my question in the placeholder element? And in subsequent queries, send the question + answer history as matching separate elements?

5

@Hussain_Akbar Yes, that is correct.

Yes, for a fresh query, you can remove the existing elements and use a single types.Content element to start with.
For subsequent queries, send the entire chat history (previous questions and answers) in separate elements to provide context.

6

The roles identify the matching query+answer pairs and the last one being a role=“user” without a matching role=“model” identifies it as a new question?

8

Yes, that’s right.

The role="user" and role="model" pairs identify the query and its corresponding answer, forming the conversational history.