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

# Configure Modal Resources

With Modal, hardware resources and autoscaling configuration are specified alongside your application code. Update the paraameters in this section by editing the values in `app.py` and redeploying.

When you clone the repo, the values are configured for an STT deployment in `us-west`.

```python
# modal_deepgram/app.py

@app.cls(
    image=engine_base_image.env({"DEPLOY_LABEL": DEPLOY_LABEL}),
    volumes={
        MODELS_PATH: models_vol,
        CACHE_PATH: cache_vol,
    },
    gpu="L4",
    secrets=[modal.Secret.from_name("deepgram")],
    timeout=30 * MINUTES,
    cpu=4,
    memory=32 * 1024,  # MB
    min_containers=1,
    region="us-west",
)
@modal.concurrent(target_inputs=64)
@modal.experimental.http_server(port=API_PORT, proxy_regions=["us-west"])
class DeepgramServer(DeepgramServerBase):
    ...
```

## Configure hardware

For Deepgram's hardware minimums, see [Deployment Environments → Engine](https://developers.deepgram.com/docs/self-hosted-deployment-environments#engine).

For Modal's GPU options, see [Modal: GPU](https://modal.com/docs/guide/gpu).

## Configure autoscaling

Modal automatically scales the number of Deepgram containers up and down based on per-container concurrency.

See their [Scaling Out guide](https://modal.com/docs/guide/scale) and [Input Conccurrency guide](https://modal.com/docs/guide/concurrent-inputs) for the different parameters and their functionality. Note that not all available parameters are surfaced in `app.py`.

> **Info**
>
> Deepgram recommends keeping at least one container active to ensure that lulls in traffic don't lead to queuing or 503s when scaling back up from zero. In Modal, set `min_containers = 1`.

> **Info**
>
> Web endpoints served with the `http_server` only accept a value for `target_inputs` and not `max_inputs`. This number should be set slightly below the active request limit in your `engine.toml` file (see [Auto-Scaling: Enforcing Limits](https://deepgram-preview-d6057290-1e55-43b9-8ea3-04ee19987964.docs.buildwithfern.com/docs/autoscaling-best-practices#enforcing-limits)).

## Configure regions

To optimize network latency, you will likely want to set the `PROXY_REGION` AND `SERVER_REGION` and route traffic from clients in those regions to that deployment.

* `PROXY_REGION` specifies the location of the Modal proxy that routes requests to containers. It can take one of four values: `us-east`, `us-west`, `eu-west`, `ap-south`.
* `SERVER_REGION` specifies which region(s) the server containers can reside in. See the Modal [Region Selection doc](https://modal.com/docs/guide/region-selection) for more information.

> Configure hardware, autoscaling, and region selection for a Deepgram Modal deployment.