> ## Documentation Index
> Fetch the complete documentation index at: https://docs.evidence.studio/llms.txt
> Use this file to discover all available pages before exploring further.

# Self Host on Fly.io

> Self-host Evidence on Fly.io from the dashboard or the fly CLI.

Fly.io runs the container as Firecracker VMs close to your users, with optional scale-to-zero.

Fly needs two pieces of configuration the other platforms don't: **memory** (the default 256mb machine is too small and the server gets OOM-killed at boot) and **port** (Fly routes to `internal_port` but does not provide `PORT` to the container, so the server listens on its default `3000`).

<Steps>
  <Step title="Add a Dockerfile">
    ```docker filename="Dockerfile" theme={null}
    FROM evidencedev/serve:latest
    COPY --chown=evidence:evidence . /project
    ```
  </Step>

  <Step title="Modify and commit connection.yaml">
    The project must use a [direct connector](/cli/connections), with secrets replaced by `${VAR}` environment references, for example:

    ```yaml filename="connection.yaml" theme={null}
    type: snowflake
    account: xy12345.us-east-1
    user: ${SNOWFLAKE_USER}
    private_key: ${SNOWFLAKE_PRIVATE_KEY}
    warehouse: COMPUTE_WH
    database: ANALYTICS
    ```

    `connection.yaml` is gitignored by default. Once its secrets are `${VAR}` placeholders, remove it from `.gitignore` and commit it — the container reads it at runtime.
  </Step>

  <Step title="Deploy">
    <Tabs>
      <Tab title="Using the Fly UI">
        In the [dashboard](https://fly.io/dashboard), click **Launch an App**, choose **Launch an App from GitHub**, and select the project's repo, granting Fly access if prompted.

        Click **Customize deploy** and set:

        * **Internal port**: `3000`
        * **Memory**: at least 512MB memory
        * **Environment Variables**: `EVIDENCE_BASIC_USER`, `EVIDENCE_BASIC_PASSWORD`, and every `${VAR}` referenced in `connection.yaml`, here `SNOWFLAKE_USER` and `SNOWFLAKE_PRIVATE_KEY`

        Deploy, then optionally enable **Auto Deploy** under **Deployments → Settings** so pushes to the default branch redeploy automatically.
      </Tab>

      <Tab title="Using the Fly CLI">
        Launch without deploying to generate `fly.toml`:

        ```bash theme={null}
        fly launch --no-deploy
        ```

        Edit `fly.toml` so the port and memory are right:

        ```toml filename="fly.toml" theme={null}
        [http_service]
          internal_port = 3000   # the port the image listens on

        [[vm]]
          memory = '512mb'       # or more - default 256mb OOM-kills the server
        ```

        <Warning>
          Use a single memory key. `fly launch` scaffolds sometimes include `memory_mb` — if both `memory` and `memory_mb` are present with different values, the wrong one can win.
        </Warning>

        Set the secrets:

        ```bash theme={null}
        fly secrets set \
          SNOWFLAKE_USER=... \
          SNOWFLAKE_PRIVATE_KEY=... \
          EVIDENCE_BASIC_USER=reports \
          EVIDENCE_BASIC_PASSWORD=changeme
        ```

        Then deploy:

        ```bash theme={null}
        fly deploy
        ```

        Healthy logs show `Ready at http://localhost:3000`.
      </Tab>
    </Tabs>
  </Step>
</Steps>

## Notes

* Ensure fly has enough memory or else you may get `Out of memory: Killed process (evidence)` in `fly logs`. Increase memory with `fly scale memory 512` or in `fly.toml`.