This URL is not the founder portal and not a founder URL. Global provider and model defaults live on founder.workweaver.ai → Models. Tenant embedding overrides live in Settings → Models.
Set the embedding model for each purpose on a specific tenant, or use
__platform__ for the global default. Setting a new model for an
in-use tenant queues a dual-write migration (#3210 companion).
| Purpose | Override provider |
|---|
PUT /api/v1/founder/embedding/policy/{tenant_id}),
CLI (ww admin embedding policy set),
MCP (ww.admin.embedding.policy_set).
All emit a embedding_policy_changed Decision Trace.
Set the memory-search re-ranker for a tenant, or use
__platform__ for the global default that every tenant
inherits until they choose their own in Settings → Models → Memory search.
PUT /api/v1/founder/rerank/policy/{tenant_id}).
Tenant Settings inherit this default until they save their own choice.
Revert drops a tenant override; it cannot delete the platform default.
Embeddings are model-specific vectors that power semantic search. Changing the embedding model requires migrating all existing vectors through a careful dual-write lifecycle:
Capability matrix (modalities + dream Purpose models):
GET /api/v1/founder/embedding/capability-matrix.
Multimodal BYOI honesty: text-only providers never stamp image/audio on a text vector.
Live available_providers for the platform default:
Supported models:
amazon.titan-embed-text-v2:0 — Titan text (1024-dim, AWS-native)amazon.nova-2-multimodal-embeddings-v1:0 — Nova multimodal (1024-dim, AWS-native, image+text)cohere.embed-english-v3 — Cohere v3 (1024-dim, AWS-native)gemini-embedding-002 — Gemini Embedding 2 (768-dim, Google AI, multimodal incl. audio)text-embedding-3-large — OpenAI 3-large (3072-dim, text-only, parity with gbrain default)text-embedding-3-small — OpenAI 3-small (1536-dim, text-only, cheaper alternative)