Models

This page used to describe a settings tab that has been removed

Pre-2026-04 versions of Ishvana shipped a Settings → Models tab where you entered Anthropic / OpenRouter / Ollama credentials, picked which model handled which tier of work, configured per-agent overrides, and tuned context-compression weights. That tab no longer exists. The Divinity Engine replaced every external provider with bundled local handlers, so there’s nothing to configure.

If you arrived from an older link, here’s what’s true now:

• No API keys to enter. Every analysis Ishvana runs is local. No request goes to Anthropic, OpenRouter, or any other cloud service.
• No model tiers to assign. There aren’t multiple models. The Divinity Engine ships one bundled embedding model (all-mpnet-base-v2 ONNX), one NLI model (nli-deberta-v3-xsmall ONNX), and an optional Deep Analysis Pack that adds spaCy + FastCoref on demand.
• No per-agent overrides. Each agent maps to a fixed set of Divinity Engine handlers documented in Etherforce overview → Handler registry.
• No costs to track. Zero per-token billing because zero tokens are billed.

Settings overview What's actually configurable in Ishvana today — and what isn't.

Etherforce overview The dispatch layer that replaced the LLM-routing layer.

Engine overview How the Divinity Engine fits together.

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