Langflow agent default current date tool causes excessive token consumption due to embedded enums and timezones

[MustafaDUT]

Bug Description

The default current date tool in Langflow agents automatically injects all enums (e.g. country codes, timezones) into the system prompt. This leads to unnecessary token consumption of around 4600 tokens per run. When an external current date component is also connected, the token usage increases dramatically to around 9600 tokens. This bloats the system prompt, increases cost, and degrades agent decision quality.

Reproduction

1️⃣ Create an agent with any basic flow in Langflow.
2️⃣ Observe that the agent prompt includes all enum values and timezone data from the default current date tool.
3️⃣ Add an explicit external current date component to the flow.
4️⃣ Token usage in the system prompt significantly increases, leading to inefficient processing.

Expected behavior

➡ The current date tool or component should only inject essential data (such as the actual current date and selected timezone).
➡ Full enum and timezone lists should not be part of the prompt by default.
➡ Token consumption should be optimized to avoid unnecessary overhead.

Who can help?

No response

Operating System

Macos Sequoia 15.4.1

Langflow Version

1.4.3

Python Version

3.11

Screenshot

As you can see on the screen, get_current automatically generates 4.5k tokens without doing anything.

When I add the external history tool it looks like this.

Flow File

No response

[MustafaDUT]

I would like to add this.

When you turn off the current date from the agent settings, yes, the token is deducted. This is a clear answer. However, what I mean here is not to reduce the agent's token consumption. It is to reduce the amount of token consumed by the CurrentDate component.

[dosubot]

The high token usage from the CurrentDate component happens because all timezone options are serialized into the tool's schema and sent to the LLM, which bloats the prompt size see code. There isn't a config or setting to limit this—it's a direct result of how DropdownInput options are handled. The only way to reduce the token consumption is to change the component code to trim or filter the timezone list, or add logic to avoid injecting the full enum list into the schema. If you want pointers on where in the code this happens or ideas for a workaround, let me know!

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}   

[MustafaDUT]

Yes, I have examined the code.
I am using only a few that I want as default as you suggested. But people may be using this consumption without seeing it. For this reason, a solution should be provided.

It is very boring for AI to respond to everything. I hope a dashboard is being followed for these responses.

[edwinjosechittilappilly]

@ogabrielluiz @rodrigosnader

[edwinjosechittilappilly]

The agent component have an option to disable the current date tool, I hope this would enable you to solve this issue.

[MustafaDUT]

Hi. Yes, there is an option to disable the default date in the controller section. But when used, it increases the prompt by 4500 tokens. We need to solve this.