Why Gemini is the ultimate embedded brain

Most products call an LLM the way they call a weather API: request in, string out, move on. That works until the model needs context the rest of your system already has.

Core, not corner

In Botagoz, Gemini sits at the center of the request lifecycle. It receives normalized geo-context from the Antigravity bus and returns structured intent that the pipeline can execute deterministically.

const intent = await gemini.reason({
  query: userText,
  context: bus.snapshot('geo'),
  schema: GeoIntentSchema,
})

const places = await pipeline.resolve(intent)

The model reasons. The pipeline verifies. The user gets an honest answer.