Write something.
A warm, private writing companion. Inference runs on your device — nothing is sent to a server.
No model loaded.
OAICA
OAICA
Browser models download once and cache offline; Prism uses the local wasm engine and requires uploaded files.
Upload a small .safetensors checkpoint (v1 minimal path) or a full model (.safetensors + config.json).
Per-request adapter, applied only to your own calls — doesn't affect other users.
Upload adapter_config.json + adapter_weights.json, then select and attach. LoRA is only supported by the Prism wasm path.
A warm, private writing companion. Inference runs on your device — nothing is sent to a server.
No model loaded.