No thinking tokens/step in response from flash-thinking model

[amrit073]

Environment details

• OS type and version: Arch linux
• Python version: 3.10.12
• pip version: 23.0.1
• google-cloud-aiplatform version: 1.78.0

Code example

    async def simple_llm_call():
        client = GenerativeModel(model_name="gemini-2.0-flash-thinking-exp-1219")
        res = await client.generate_content_async(
            stream=True,
            generation_config={"max_output_tokens": 10000},
            contents=Part.from_text("Solve 3*x^3-5*x=1"),
        )
        async for chunk in res:
            print(chunk.candidates[0].content)

    asyncio.run(simple_llm_call())

Output

role: "model"
parts {
  text: "We want to solve the equation $3x^3 - 5x = 1$,"
}

role: "model"
parts {
  text: " which can be rewritten as $3x^3 - 5x - 1 = 0$.\nThis is a cubic equation of the form $ax^3 + bx + c = 0$, where $a=3$, $b=-5$, and $c=-1$.\nWe use the trigonometric"
}

role: "model"
parts {
  text: " method for solving cubic equations of this form.\nLet $x = k \\cos \\theta$. Substituting this into the equation, we get:\n$3(k \\cos \\theta)^3 - 5(k \\cos \\theta) - 1 = 0$\n$3k^3 
\\cos"
}

Acc to docs and linked notebook ,
The first part should include the model's thoughts instead of an answer. And it would also have part[0].thought True but it is not showing up in our case.