Embeddings
Swap the WorkMemory embedding model and rebuild the local pgvector index.
Back
Active
Current embedding
Provider
—
Model
—
Dimension
—
Swap
Pick a new embedding provider
Provider
bedrock-titan (Amazon Titan Embed V2, 1024-dim)
ollama (local Ollama server)
sentence-transformers (local model, lazy import)
huggingface (HuggingFace Inference API)
openai-compatible (OpenAI / Azure / Together / vLLM)
voyage (Voyage AI)
cohere (Cohere Embed v4)
jina (Jina embeddings v3)
Model
Base URL (Ollama / OpenAI-compatible only)
API key (paste — never logged)
Swap & rebuild
Warning:
Swapping the embedding provider re-embeds every stored WorkMemory fact. For Ollama you must first
ollama pull <model>
. Embedding dimension may change (Titan = 1024, nomic-embed-text = 768).
Job
Rebuild progress
Job ID
—
State
—
Progress
0%