llm-sidecar

Settings

Without a key the sidecar is local-only. With one it can also route to OpenRouter's free models — verified working before each use, and rotated when one throttles.

checking…
Budget

The key is applied immediately and never sent back to this page — only a masked preview. Unless you tick "remember", it lives in the running daemon and is gone when you restart it.

Status

Loading…

Search

Loading…

Cache

Loading…

Local models — will they fit?

Normally a tier picks its own model each time, testing candidates until one answers. Clicking a slot below locks that tier to one model instead: no searching, no budget, no swapping if it breaks. Click the same slot again to unlock it.

ModelSizeNeeds ParamsQuantVerdictAlways use for…
Loading…

Sized for an 8k context, with 3 GiB left for the rest of the machine. Locks apply to this running daemon only — they are not written to your config, and a locked tier ignores the budget entirely.

Usage — last 30 days

Loading…
today
ModelCallsCost TokensCachedAvg