Skip to content

Provider routing matrix

Providers

3

Routing profiles

4

Default profile

gracekelly-primary

IDLabelKindEnabledAuth env varModels
ollamaOllamalocalenabled2
gracekellyGraceKelly local orchestratorlocalenabledGRACEKELLY_API_KEY3
mistralMistralpaidenabledMISTRAL_API_KEY4

Prices are USD per 1M tokens.

ProviderModelAliasesInputOutput
ollamaqwen2.5:7bollama-small, ollama-fast$0.00$0.00
ollamallama3.2:3bollama-tiny$0.00$0.00
gracekellysonar-2gk-sonar, gk-fast$0.00$0.00
gracekellyclaude-sonnet-4-6gk-claude-sonnet, gk-strong, claude-sonnet-4-6-api$0.00$0.00
gracekellygpt-5-4-apigk-gpt-5$0.00$0.00
mistralministral-3b-latestministral-3b, mistral-tiny$0.04$0.04
mistralministral-8b-latestministral-8b$0.10$0.10
mistralmistral-small-latestmistral-small$0.20$0.60
mistralmistral-large-latestmistral-large$2.00$6.00

The default profile is marked (default).

ProfileFast tierStrong tierFallbackDescription
local-firstollama/qwen2.5:7bollama/qwen2.5:7bExplicit local-only Ollama routing, zero paid spend.
gracekelly-primary (default)gracekelly/sonar-2gracekelly/claude-sonnet-4-6ollama/qwen2.5:7bGraceKelly orchestrator for both tiers (Perplexity Pro-backed default), with explicit Ollama fallback on failure.
external-mistralmistral/ministral-3b-latestmistral/mistral-small-latestDirect Mistral API for external deployments without GraceKelly; cheap, dependable, no local orchestrator dependency.
gracekelly-mixedmistral/ministral-3b-latestgracekelly/claude-sonnet-4-6Mixed routing — Mistral API for fast tier (classify/transform/grade_docs/verify_facts/extract_claims/online_evaluators), GraceKelly browser for strong tier (final answer + suggest_questions). Reduces browser submits per case from 4-7 to ~3 while keeping full Self-RAG / Corrective RAG / auto-route intact. Also a valid production routing for single-user local deploys with both Mistral key and GraceKelly available.