Payload Optimization

Practical ways to cut tokens, latency and cost using fields you already have.

#Principles

Reuse a stable conversationId for follow-up turns instead of restating context — see Multi-turn.
Cache deterministic results (charts, panchang) client-side — they do not change for the same inputs.
Choose Vedika Eco for high-volume, cost-sensitive calls — see Choosing a model.
Keep questions focused. Answer length adapts to the question, so a tighter question costs less.
Request only the language you need, and set responseFormat to the shape you will actually consume so you don't re-parse it.
Batch independent work with the Batch API instead of firing many single calls.

Field	Effect
`conversationId`	Reuse it to avoid restating birthDetails and context on every follow-up.
`speed`	Set `"eco"` for high-volume, cost-sensitive workloads.
`responseFormat`	Request the shape you consume — `json` or `markdown` — to skip re-parsing.
`language`	Request only the language you need.
`question`	Tighter, more focused questions produce shorter, cheaper answers.