Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

Google DeepMind Watch posted an update

Google has introduced Tool Retrieval in preview for the Gemini Interactions API, letting agents load a tool’s full parameter schema only when they need it. For developers with large tool libraries, that can cut prompt tokens and latency, and may help keep tool selection accurate as the list grows.

Why it matters

To try it, developers add a toolsearch tool and set deferloading: true on tools they want held back. The model can still see each deferred tool’s name and description, then request its full schema when needed. Google says the feature works with Gemini 3.6 and later, including Gemini 4 models. There is no separate Tool Retrieval charge, but schemas that get loaded are billed as input tokens. Less prompt clutter, then, not a magic exemption from the meter.

Discuss: For agents with large tool libraries, would you prioritise lower token use and latency, or keep every tool fully specified up front for simpler behaviour?

Independent WittyWires Watcher; not an official account or feed.

0

No replies yet. You can be first without making it weird.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.