feat(fusion): parameter sweep for doctor validation

.github/workflows/test.yml

@@ -33,13 +33,28 @@ jobs:
       - name: Run Prettier --check
         run: prettier --check .
 
+  python-tests:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v3
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@v3
+
+      - name: Run generate_report tests
+        run: uv run --no-project --with "pytest==9.0.2" --with "jinja2==3.1.6" --with "humanize==4.15.0" --with "markupsafe" python -m pytest tests/ -v
+
   nf-test:
     runs-on: ubuntu-latest
     env:
       NXF_ANSI_LOG: false
     steps:
       - uses: actions/checkout@v3
 
+      - name: Install uv
+        uses: astral-sh/setup-uv@v3
+
       - uses: actions/setup-java@v3
         with:
           distribution: "temurin"

.prettierignore

@@ -13,4 +13,5 @@ results/
 *.test
 *.test.snap
 *.txt
+assets/templates/
 

README.md

@@ -143,25 +143,65 @@ This process tests the ability to use a GPU. It uses the `pytorch` conda environ
 
 _Note: Enabled only if the parameter `--fusion` is specified (or set to `true` by a profile)._
 
-This process runs `fusion doctor` to validate the Fusion filesystem configuration. It checks system requirements (kernel version, memory, and disk space), FUSE device availability, and cloud bucket accessibility. The process produces a JSON diagnostic report published to `${outdir}/fusion/`.
-
-This test requires the `fusion` binary to be available in the task environment and will fail if it is not found.
+This process runs the `fusion doctor` diagnostics tool to validate the Fusion filesystem configuration. It checks system requirements (e.g. kernel version, memory, disk space, etc.), FUSE device availability, and cloud bucket accessibility among others. The process produces a JSON diagnostic report published to `${outdir}/fusion/`.
 
 #### Fusion Profiles
 
-Use a built-in profile to enable Fusion validation with recommended thresholds:
+Use a built-in profile to enable Fusion validation with predefined thresholds:
 
 ```bash
 nextflow run seqeralabs/nf-canary -profile fusion_aws_recommended
 ```
 
-Available profiles:
+> [!TIP]
+>
+> **If in doubt, start with the `recommended` tier:**
+>
+> 1. Choose `low` for small workloads or quick smoke tests, and `high` for large-scale production with big datasets.
+> 2. Choose `recommended` to match Seqera's documented minimum requirements for production workloads with Fusion.
+> 3. Choose `high` (or a custom threshold of 400 GB+ storage) if your pipeline processes files larger than 100 GB.
+
+##### AWS
+
+> [!NOTE]
+>
+> Seqera Platform will auto-select NVMe-based instance families when Fusion is enabled (e.g. `m6id`, `c6id`, `r6id`) and the "Fast instance storage" toggle is active in the CE settings. Fusion can also work without NVMe instances, but in this case the EBS disk shall be bumped to 100 GB (`gp3`, 325 MB/s); this is what the `low` profile validates.
+
+| Profile                  | Disk   | Memory | Kernel | Based on            |
+| ------------------------ | ------ | ------ | ------ | ------------------- |
+| `fusion_aws_low`         | 100 GB | 4 GB   | 5.10+  | EBS gp3 (no NVMe)   |
+| `fusion_aws_recommended` | 200 GB | 8 GB   | 5.10+  | Seqera docs minimum |
+| `fusion_aws_high`        | 474 GB | 16 GB  | 5.10+  | m6id.2xlarge NVMe   |
+
+**Google Cloud**
+
+> [!NOTE]
+>
+> Seqera Platform auto-selects families supporting local SSDs (e.g. `n2`, `c2`, `n2d`) and provisions a 375 GB NVMe SSD per job.
+
+| Profile                     | Disk   | Memory | Kernel | Based on            |
+| --------------------------- | ------ | ------ | ------ | ------------------- |
+| `fusion_google_low`         | 50 GB  | 4 GB   | 5.15+  | GCP persistent disk |
+| `fusion_google_recommended` | 375 GB | 8 GB   | 5.15+  | 1x local NVMe SSD   |
+| `fusion_google_high`        | 750 GB | 16 GB  | 5.15+  | 2x local NVMe SSDs  |
+
+##### Azure
+
+> [!note]
+>
+> Unlike AWS and Google Cloud, there is no auto-selection: the user must pick the VM size. Seqera recommends E-series with a `d` suffix (e.g. `Standard_E8d_v5`, `Standard_E16d_v5`).
+
+| Profile                    | Disk   | Memory | Kernel | Based on                     |
+| -------------------------- | ------ | ------ | ------ | ---------------------------- |
+| `fusion_azure_low`         | 75 GB  | 4 GB   | 5.15+  | `Standard_E2d_v5` temp disk  |
+| `fusion_azure_recommended` | 300 GB | 8 GB   | 5.15+  | `Standard_E8d_v5` temp disk  |
+| `fusion_azure_high`        | 600 GB | 16 GB  | 5.15+  | `Standard_E16d_v5` temp disk |
+
+#### Caveats
+
+In AWS Batch, the Seqera Platform UI allows selecting instance families (e.g. `m6id`) but not specific sizes. Small instances in an otherwise valid family may not meet the `recommended` disk threshold. For example, on AWS the `m6id` family NVMe ranges from 118 GB (`.large`) to 1,900 GB (`.8xlarge`), with only `.xlarge` and above meeting the 200 GB threshold.
 
-| Profile                     | Description                               |
-| --------------------------- | ----------------------------------------- |
-| `fusion_aws_recommended`    | AWS Batch recommended thresholds          |
-| `fusion_google_recommended` | Google Cloud Batch recommended thresholds |
-| `fusion_azure_recommended`  | Azure Batch recommended thresholds        |
+If `fusion doctor` reports a disk requirement failure, request more CPUs/memory to get a larger instance, or use the `low` profile for small tasks.
 
 #### Custom Requirements