Back to Documentation
Spoke Modes: Lite (8.7MB) vs Full (~120MB)
Choosing between zero-dependency light mode and air-gapped embedded models.
LOB Spoke offers two configurations optimized for local hardware constraints.
LOB Spoke Lite (8.7MB - Default):
Employs SQLite FTS5 for rapid full-text search. Defers semantic vector tasks to the Master Hub. This keeps the Spoke's memory footprint under 20MB of RAM.
LOB Spoke Full (~120MB - Air-gapped):
Bundles the ONNX runtime and a lightweight embedding model (e.g., all-MiniLM-L6-v2) directly inside the binary. Computes embeddings locally for absolute privacy in air-gapped environments.