KernelLoom

Local runtime control

Paths entered here are resolved by the server machine. Settings apply to the current process and are not written to disk unless you download a configuration file.

Resident models

Warm, clear caches, reuse settings, or unload a model.

0 models

Chat

Uses the OpenAI-style chat endpoint. Streaming renders each received token.

Hardware

Inspect local devices and the runtimes that report them.

No hardware profile loaded.

RAG collections

Create a local collection, ingest server-local paths or inline text, then retrieve or ask against it.

1. Create a collection

Advanced retrieval and prompt settings

2. Ingest knowledge

Text that matches an existing server path is loaded from disk. Use the API key and server access controls for networked deployments.

3. Retrieve or ask

Active collections

Removing a collection releases only its store; resident models remain loaded.