- Ollama running on the user’s computer
- A downloaded local model
- DecaCap Local Companion
- Enough RAM or unified memory for the model
local runtime · paired
Ollama
Local text inference from any entitled Workbench box through the user-paired loopback companion.
- DecaCap interface
- Device-local /api/chat and /api/tags
- Published evidence
- Companion pairing, authenticated health, model listing, chat routing, CORS boundary, pair capture, and lifecycle endpoints are implemented. Actual model speed and fit depend on the user’s device.
- Exact-model probe
- Not published Account-, model-, region-, quota-, runtime-, and hardware-specific verification still requires a real user-owned key or device.
- Last reviewed
- 2026-07-26
- The website never installs Ollama or a model
- A local model is not automatically trained by using it
- Model and hardware compatibility must be checked by the user
- Authenticated loopback health
- Model discovery
- Local chat
- No site-owned GPU
- Local data path
- Access to a particular model on a user account
- Quality, speed, pricing, quota, or safety behavior
- Hardware fit or training-toolkit compatibility