Models

Platform inference defaults

What every new workspace gets before anyone touches Settings. Publishes persist platform-wide; a Decision Trace event is recorded per publish.

Cloud curator

Primary and fallback lists for Workweaver Cloud. The menu is the Bedrock catalog from the founder providers API. Bedrock serves these today; nothing here blocks a later mixture. An empty store stays empty until you choose a model.

Search catalog

Primary models

Fallback models

HTTP POST /api/v1/founder/inference/defaults · CLI ww admin inference defaults.set · MCP ww.admin.inference.defaults_set

Ordered fallbacks

Primary provider, then the exact order to try on failure. Published per capability class — never seat labels.

Primary provider

Fallback order (first tried first)

HTTP POST /api/v1/founder/inference/routing · CLI ww inference routing.set · MCP ww.inference.routing.set

Embedding default

Platform-wide (__platform__) embedding policy: aws_only · google_enabled · dual_provider. Primary and fallback embedding models come from the same Bedrock catalog as chat — only rows the embedding policy accepts are offered.

Mode

Index dimension (optional guard)

Embedding primary models

Embedding fallback models

HTTP PUT /api/v1/founder/embedding/policy/__platform__ · CLI ww admin embedding policy.set · MCP ww.admin.embedding.policy_set

Memory search — re-ranker default

Platform-wide (__platform__) re-ranker for memory search. Tenants may override under Settings → Models → Memory. Choices come from the founder re-rank policy catalog — not a hardcoded list.

Re-ranker

HTTP PUT /api/v1/founder/rerank/policy/__platform__ · No CLI twin yet · MCP ww.memory.rerank_settings

Voice defaults

Platform voice row: per-minute ceiling and consent defaults. Optimistic locking — conflicts are surfaced, never overwritten.

Minutes ceiling / month

HTTP PUT /api/v1/founder/voice/defaults · CLI ww admin voice defaults set · MCP ww.voice.admin.defaults_set

Provider catalog

Every configured platform provider. Test runs a real smoke-test call — the result and any error are copyable.

Provider	State	Default model	Health	Test
Loading provider catalog…

HTTP POST /api/v1/founder/inference/providers/{id}/smoke-test · CLI ww inference provider.test · founder-authenticated (no MCP surface — #17353)

Workspace effective config

Read-only: what a workspace actually resolves right now (platform policy + override). No editing here.

Workspace id or slug

Provider connections

xAI SuperGrok stays on OAuth. Anthropic Claude connects by API key — no CLI DirectSDK. ChatGPT and Codex subscriptions are allowed.