Could you clarify if Google.genai has memory or is stateless?

[XinyueZ]

Hey master,

Will the result of auto function call (AutomaticFunctionCallingConfig) be remembered?
I just wan to know yes or no.

thank you

[Venkaiahbabuneelam]

Hi @XinyueZ, Thanks for reaching out us!

As per our documentation, Google.genai is stateless by default, unless you explicitly manage and provide conversation history or past function call results in your application, the client-API does not automatically remember previous outputs from AutomaticFunctionCallingConfig.

Thanks

[XinyueZ]

@Venkaiahbabuneelam thanks, then question continues: ignore_call_history, how to understand this flag?

In my humble opinion, generally, the model needs "context" to decide which tool to call, and then the tool's result will be appended to the end of the previous "context," forming a new "context" to continue. If all goes smoothly, the model may indicate a desire for the next tool. In short, every time the model hopes for a particular tool, it must rely on the "history" of the last interaction. Is there really no place in the SDK to maintain this "history"?

For example, let's create a simple problem involving addition, subtraction, multiplication, and division, using the tools add, sub, mul, and div. How is it possible to choose to call the next tool when we do not know the result of the previous tool's step?

For instance, on Ollama, we at least need to maintain a loop （recursion can also be, for example, langgraph) to automatically and "continuously" call tools, constantly providing "context" to the model, at least in "memory," right?

For example, this logic:

https://github.com/XinyueZ/ollama-python/blob/983c51ea082597f22b468eb1c9e7807ca9561b0a/examples/ollama_chat_automatic_function_calling.py

If I guess correctly:
The model shouldn't know the results of the "before before" tool call, right?

[pact4bin]

This platform is too complex for me and I need help with understanding what to do

…

On Wed, 18 Feb 2026, 2:04 am ChrisDEV, ***@***.***> wrote: *XinyueZ* left a comment (googleapis/python-genai#2046) <#2046 (comment)> @XinyueZ <https://github.com/XinyueZ> thanks, then question continues: ignore_call_history, how to understand this flag? — Reply to this email directly, view it on GitHub <#2046 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BQXVSIBZPZERX37TP2UFSXT4MMUZJAVCNFSM6AAAAACURNVNXOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTSMJVGI2DMNRZHA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[Venkaiahbabuneelam]

Hi @XinyueZ,
Thank you for your feedback. We appreciate you taking the time to share your thoughts with us

Below is the actual behavior depends on whether you are looking at the API level or the SDK level:

The Gemini API is Stateless: By default, the underlying REST RPC API does not store conversation history. Every request to generate_content is independent. If you want the model to "remember" previous turns, you must provide the entire conversation history in the contents list of each new request.

The google-genai SDK provides Stateful Abstractions: The Python SDK includes a ChatSession class specifically to manage this state for you on the client side. When you use chat.send_message(), the SDK automatically appends the messages to an internal history buffer and sends that accumulated history to the API.

Below is the url

Thanks

[Venkaiahbabuneelam]

Create a chat session to start a multi-turn conversations with the model. Then, use chat.send_message function multiple times within the same chat session so that it can reflect on its previous responses (i.e., engage in an ongoing conversation). See the ‘Create a client’ section above to initialize a client.

"chat = client.chats.create(model='gemini-2.5-flash')
response = chat.send_message('tell me a story')
print(response.text)
response = chat.send_message('summarize the story you told me in 1 sentence')
print(response.text)"

[github-actions]

This issue has been marked as stale because it has been open for 7 days with no activity. It will be closed in 2 days if no further activity occurs.