sceptre-managing-models

<!-- AI-RULEZ :: GENERATED FILE — DO NOT EDIT Content-Hash: blake3:8c96dd895d636cc85419c33a4767457ad757165f29e9554be0481b89d1220af0 Source-Hash: blake3:5c8442fa8b2f9f7d5d68801c4562f6bf03cc6e52237900bbcf720e4b5bedfa5e Schema-Version: v1 --> # Managing Models Run `sceptre models list --all` to discover models and their cache status. Download only the required languages with `sceptre models download --lang english --lang korean`; use `--all` only when all language models are needed. Do not combine `--all` and `--lang`. Models use the Hugging Face hub cache configured by `HF_HUB_CACHE` or `HF_HOME`. Inspect the chosen runtime with `sceptre env --format json`. The CLI defaults to ONNX Runtime and downloading. A pure Rust installation is `cargo install sceptre-cli --no-default-features --features tract,download`; select `--backend tract` when running it. MCP support requires the additional `mcp` feature. Do not assume an installed CLI includes MCP. For offline library use, provide verified model assets or explicit detector and recognizer paths; do not disable checksum validation to work around a failed model download.