> This page is for NVIDIA NIM for Batched Geometry Relaxation, version 1.1.0 (Latest) (default).
> For other versions, use one of these documentation indexes:
> - 1.1.0 (Latest) (default): https://docs.nvidia.com/nim/alchemi/alchemi-bgr/1.1.0/llms.txt
> - 1.0.0: https://docs.nvidia.com/nim/alchemi/alchemi-bgr/1.0.0/llms.txt

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/nim/alchemi/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nim/alchemi/_mcp/server.

# Environment Variables

Configure NIM for BGR behavior at container startup using the
following environment variables.

## Required Variables

The following variable is required for authenticated access to the NGC
container registry:

| Variable | Required | Default | Description |
|---|---|---|---|
| `NGC_API_KEY` | Yes (for NGC access) | -- | NGC authentication key used to access the NIM container. The default MACE-MPA-0 checkpoint is downloaded separately from Hugging Face. |

## Configuration Variables

Use the following variables to configure the NIM:

| Variable | Type | Default | Description |
|---|---|---|---|
| `ALCHEMI_NIM_MODEL_TYPE` | string | `"mace"` | Model architecture: `mace`, `aimnet2`, or `tensornet`. Use `ALCHEMI_NIM_MODEL_PATH` to select the checkpoint. |
| `ALCHEMI_NIM_MODEL_PATH` | string | null | Path to the model file or directory inside the container. |
| `ALCHEMI_NIM_PBC` | bool | true | Backend PBC mode. When true, every request structure must include `cell`; when false, every structure must omit `cell`. Nested `atoms[].pbc` values do not override this mode. |
| `ALCHEMI_NIM_BATCH_SIZE` | int | null | Fixed batch size (`null` enables automatic estimation at startup). |
| `ALCHEMI_NIM_BATCH_SIZE_ESTIMATE_DATA` | string | `"omat24"` | Dataset used for automatic batch-size estimation: `omat24`, `omol25`, or `omc25`. |
| `ALCHEMI_NIM_BGR_OPTTOL` | float | 0.005 | Force tolerance (eV/Å). Can be overridden per request. |
| `ALCHEMI_NIM_BGR_MAX_STEPS` | int | 2000 | Backend-wide maximum optimization steps, set at startup. |
| `ALCHEMI_NIM_BGR_CELLOPT` | bool | false | Enable cell optimization. Can be overridden per request. |
| `ALCHEMI_NIM_BGR_OPTTOL_PRESSURE` | float | 0.5 | Pressure tolerance (kBar) for cell optimization. Can be overridden per request. |
| `ALCHEMI_NIM_BGR_OPTIMIZER_PRESET` | string | `"materials"` | Optimizer preset: `materials` or `molecules`. |
| `ALCHEMI_NIM_DFTD3_ENABLED` | bool | false | Enable DFT-D3(BJ) dispersion correction. |
| `ALCHEMI_NIM_DFTD3_PARAM` | string (JSON) | null | JSON string containing custom DFT-D3(BJ) parameters. Custom models require `a1`, `a2`, and `s8`. Refer to [Configuration](/nim/alchemi/alchemi-bgr/configuration#dft-d3-parameters) for details. |

## Container Runtime Variables

Use the following variables to configure the container runtime:

| Variable | Required | Default | Description |
|---|---|---|---|
| `NIM_HTTP_API_PORT` | No | `8000` | HTTP port for the inference API. |
| `ALCHEMI_NIM_DISABLE_MODEL_DOWNLOAD` | No | false | Whether to disable the automatic startup download of the default MACE-MPA-0 checkpoint from Hugging Face. Set to `true` when mounting a user-provided model. |

For launch examples, refer to [Quickstart Guide](/nim/alchemi/alchemi-bgr/quickstart-guide),
[Custom Models](/nim/alchemi/alchemi-bgr/custom-models), and
[Tuning Parameters](/nim/alchemi/alchemi-bgr/tuning-parameters).

## Related Topics

- [Configuration](/nim/alchemi/alchemi-bgr/configuration) for configuration options, per-request overrides, and model configuration
- [Supported Models](/nim/alchemi/alchemi-bgr/overview#supported-models) for model sourcing
- [Quickstart Guide](/nim/alchemi/alchemi-bgr/quickstart-guide) for initial launch steps