Providers & Models

This content describes a system that no longer exists

Ishvana used to route to external LLM providers (Anthropic, OpenRouter, Ollama). That layer was deleted in 2026-04 when the Divinity Engine replaced every external provider with bundled local handlers. There are no providers to configure, no model tiers to assign, and no per-agent overrides to manage — everything runs locally on deterministic handlers.

If you arrived here from an older link, the pages that replaced this one are:

Etherforce overview The dispatch layer that used to route LLM calls and now dispatches Divinity Engine handler invocations. Same name, very different job.

Engine overview The mental model for how the Divinity Engine fits together with the rest of the backend.

Etherforce Observability Live diagnostics for handler dispatches — the surface that replaced the old per-provider cost-tracking views.

What changed

Pre-2026-04 versions of Ishvana shipped an external-provider routing layer. You signed up for an Anthropic key, an OpenRouter key, or you stood up a local Ollama server, and the app routed agent requests to whichever provider you’d configured for that tier of work. That layer is gone. The Divinity Engine — a stack of bundled deterministic handlers covering consistency checking, prose analysis, lore graph work, generative replacements, and classifier work — replaced every external provider in one cycle.

The practical effect: there’s no API key to enter anywhere in Settings, no model assignments to manage, no token cost to budget against. Every agent feature in Ishvana runs on local handlers. Costs are zero. Nothing leaves your machine. Same input always gives the same output. The tradeoffs are different from what the original LLM-routing design described, and they’re spelled out in the Etherforce overview and the Engine overview.

This page stays here as a redirect rather than getting deleted because external links and older browser tabs still point at it.