Model Management

Manage embedding models for semantic search

2 min read

Model Management

The knowns model command provides full control over embedding models used for semantic search. Models are stored globally at ~/.knowns/models/ and shared across all projects.

Commands

Status

knowns model           # Show current model status
knowns model status    # Detailed status with disk usage

List Models

knowns model list      # List all available models
knowns model ls        # Alias

Download

knowns model download <model-id>
knowns model dl <model-id>    # Alias

Set Model

knowns model set <model-id>
knowns search --reindex        # Rebuild index after changing model

Built-in Models

Model IDQualityDimensionsSizeBest For
gte-smallBalanced384~50MBMost projects (recommended)
all-MiniLM-L6-v2Fast384~45MBLarge codebases
gte-baseQuality768~110MBHigh accuracy
bge-small-en-v1.5Balanced384~50MBEnglish text
bge-base-en-v1.5Quality768~110MBEnglish, high quality
e5-small-v2Balanced384~50MBGeneral use

Custom Models

Add any HuggingFace feature-extraction ONNX model:

# Add a custom model
knowns model add Xenova/bge-large-en-v1.5 --dims 1024 --tokens 512
 
# Download and use it
knowns model download bge-large-en-v1.5
knowns model set bge-large-en-v1.5
knowns search --reindex
OptionDescription
--dims <number>Embedding dimensions (default: 384)
--tokens <number>Max input tokens (default: 512)
--name <name>Display name for the model

Remove Custom Model

knowns model remove <model-id>
knowns model rm <model-id>       # Alias
knowns model rm <model-id> -f    # Also delete downloaded files

Workflow Examples

knowns init
# ? Enable semantic search? Yes
# ? Select model: gte-small (recommended)
 
knowns search "authentication"    # Works immediately

Upgrade to higher quality model

knowns model list
knowns model download gte-base
knowns model set gte-base
knowns search --reindex

Add multilingual support

knowns model add Xenova/multilingual-e5-small --dims 384 --name "E5 Multilingual"
knowns model download multilingual-e5-small
knowns model set multilingual-e5-small
knowns search --reindex