Concepts
Resources vs models
Deva serves two kinds of capability through the same deva_ key. Knowing which is which tells you how to call it and how it's priced.
Deva serves two kinds of capability through the same deva_ key. Knowing which is which tells
you how to call it and how it's priced.
| Resources | Models | |
|---|---|---|
| What | Payable tools/APIs — scraping, web search, enrichment, social data | LLMs (34 frontier models) |
| Catalog | API Catalog (/app/catalog) | Models (/app/tools) |
| Call path | The resource rail: discover → inspect → estimate → run → poll | OpenAI-compatible /v1/chat/completions |
| Pricing | Per run; often variable (may need a maxTotalChargeUsd cap) | Per token |
| Result | Async run → run_id → poll to settle | Synchronous completion |
When to use which
- Need live data or an action in the world (scrape a page, search, enrich a lead)? That's a resource — find it in the API Catalog.
- Need text/reasoning/generation? That's a model — call
/v1/chat/completions.
Both draw on the same wallet (1 credit = $1.00 → Money units) and both are metered per use, idle = $0 → Pricing & settlement.