Bridge / Models
Pick the model path after the work is clear.
Model size is only one variable. Context length, concurrency, latency, hardware, tools, and data handling shape the usable system.
01
On the workstation
A local runtime can place inference beside Bridge and the authorized files, subject to the machine's capacity.
02
On a team server
Ollama or vLLM may serve a private-network endpoint. Capacity, identity, uptime, and operations become shared responsibilities.
03
External endpoints
If a provider or tool receives data, say so. A locally installed app does not make every connected service local.