Add Subparsers

Author: AryanAhadinia

README.md

@@ -1,52 +1,97 @@
 # `sml`: Swiss AI Model Launch
 
-A CLI tool for launching AI models on HPC systems via FirecREST, a remote launcher, or SLURM commands.
+A CLI tool for launching AI models on HPC systems via FirecREST or SLURM commands.
 
 ## Install
 
-1. Regular use
-   - SSH
+- SSH
 
-   ```bash
-   pip install git+ssh://git@github.com/swiss-ai/model-launch.git
-   ```
+  ```bash
+  pip install git+ssh://git@github.com/swiss-ai/model-launch.git
+  ```
 
-   - HTTPS (once public)
+- HTTPS
 
-   ```bash
-   pip install git+https://github.com/swiss-ai/model-launch.git
-   ```
+  ```bash
+  pip install git+https://github.com/swiss-ai/model-launch.git
+  ```
 
-2. Development
+## Usage
 
-   ```bash
-   git clone git@github.com:swiss-ai/model-launch.git && cd model-launch
-   uv venv --python 3.12
-   source .venv/bin/activate
-   uv pip install -e ".[dev]"
-   pre-commit install
-   ```
+### Quick Start
 
-## Usage
+Just run:
+
+```bash
+sml
+```
+
+That's it. On first run, SML will guide you through a one-time setup. After that, running `sml` again will take you straight to launching a model.
 
-### Initialization
+The rest of this section documents subcommands and CLI arguments for advanced or automated use cases.
 
-The first time you run `sml` command, you will be prompted to do the initialization, which includes providing necessary credentials and configurations.
+### Subcommands
 
-There is three ways to initialize the mean of launching commands for the CLI as listed below.
+| Command          | Description                                              |
+| ---------------- | -------------------------------------------------------- |
+| `sml init`       | Initialize SML configuration                             |
+| `sml quickstart` | Launch a model with guided prompts                       |
+| `sml advanced`   | Launch a model with advanced configuration (coming soon) |
 
-- FirecREST: You will be able to run the CLI on your local machine and the CLI will use FirecREST to submit jobs. You will need to provide your FirecREST credentials during the initialization. If you don't have the credentials, you can follow the instructions in the [Appendix](#acquiring-firecrest-credentials) to acquire them.
+### Initialization (`sml init`)
+
+Run `sml init` to configure SML. You will be prompted to provide necessary credentials and configurations.
+
+There are three ways to initialize the launcher:
+
+- FirecREST: Run the CLI on your local machine; jobs are submitted via FirecREST. You will need to provide your FirecREST credentials. If you don't have them, follow the instructions in the [Appendix](#acquiring-firecrest-credentials).
 - Remote Launcher: Not operational yet.
 - SLURM: Not operational yet.
 
-For health check, you will be prompted to provide your CSCS API key. If you don't have the API key, you can follow the instructions in the [Appendix](#acquiring-cscs-api-key) to acquire it.
+For health check, you will be prompted to provide your CSCS API key. If you don't have the API key, follow the instructions in the [Appendix](#acquiring-cscs-api-key).
+
+All prompts can be pre-filled via CLI arguments to skip interactive prompts:
+
+| Argument                       | Description                                            |
+| ------------------------------ | ------------------------------------------------------ |
+| `--launcher`                   | Job submission method (`firecrest`, `remote`, `slurm`) |
+| `--firecrest-url`              | FirecREST API URL                                      |
+| `--firecrest-token-uri`        | FirecREST token URI                                    |
+| `--firecrest-client-id`        | FirecREST client ID                                    |
+| `--firecrest-client-secret`    | FirecREST client secret                                |
+| `--remote-launcher-address`    | Remote launcher address (if using `remote`)            |
+| `--remote-launcher-auth-token` | Remote launcher auth token (if using `remote`)         |
+| `--cscs-api-key`               | CSCS API key for health checks                         |
+| `--telemetry-endpoint`         | Endpoint for telemetry reports                         |
+
+### Launching a Model (`sml quickstart`)
 
-### Launching a Model
+After completing initialization, run `sml quickstart` (or just `sml`). The CLI will guide you through selecting a model and providing the necessary launch configuration.
 
-After completing the initialization, you can simply run the `sml` command. The CLI will guide you through the process of launching a model, which includes selecting a model, providing necessary information for the launch, and confirming the launch.
+All prompts can be pre-filled via CLI arguments to skip interactive prompts:
+
+| Argument                | Description                                  |
+| ----------------------- | -------------------------------------------- |
+| `--firecrest-system`    | Target HPC system to launch on               |
+| `--firecrest-partition` | SLURM partition to use                       |
+| `--model`               | Model to launch (`<vendor>::<model>`)        |
+| `--framework`           | Inference framework to use                   |
+| `--workers`             | Number of workers                            |
+| `--use-router`          | Load balance across workers (`yes`, `no`)    |
+| `--time`                | Job time limit (`HH:MM:SS`)                  |
 
 ## Development
 
+### Setting Up Development Environment
+
+```bash
+git clone git@github.com:swiss-ai/model-launch.git && cd model-launch
+uv venv --python 3.12
+source .venv/bin/activate
+uv pip install -e ".[dev]"
+pre-commit install
+```
+
 ### Testing Environment
 
 For writing the unit tests, you have to create a `.test.sh` file in the root of the repository with the following content:
@@ -62,6 +107,8 @@ export FIRECREST_PARTITION=normal
 export CSCS_API_KEY=<your-api-key>
 ```
 
+This file will be sourced when running the tests with `make test`, and the environment variables will be available for the tests.
+
 ### Common Commands
 
 There is a `Makefile` with common development commands.
@@ -94,7 +141,7 @@ There is a `Makefile` with common development commands.
 
 ### Acquiring FirecREST Credentials
 
-Please follow the instructions in the [FirecREST documentation](https://docs.cscs.ch/services/devportal/#manage-your-applications) to acquire the necessary credentials for authentication.
+Please follow the instructions in the [FirecREST documentation](https://docs.cscs.ch/services/devportal/#manage-your-applications) to acquire the necessary credentials for authentication from [Developer Portal](https://developer.svc.cscs.ch/devportal/apis).
 
 ### Acquiring CSCS API Key
 

src/swiss_ai_model_launch/cli/main.py

@@ -129,9 +129,19 @@ def _build_parser() -> argparse.ArgumentParser:
         prog="sml",
         description="Swiss AI Model Launcher",
     )
-    InitConfig().add_to_parser(parser)
-    _make_firecrest_launcher_config().add_to_parser(parser)
-    _make_launch_request_config().add_to_parser(parser)
+    subparsers = parser.add_subparsers(dest="subcommand", required=False)
+
+    init_parser = subparsers.add_parser("init", help="Initialize SML configuration")
+    InitConfig().add_to_parser(init_parser)
+
+    quickstart_parser = subparsers.add_parser(
+        "quickstart", help="Launch a model with guided prompts"
+    )
+    _make_firecrest_launcher_config().add_to_parser(quickstart_parser)
+    _make_launch_request_config().add_to_parser(quickstart_parser)
+
+    subparsers.add_parser("advanced", help="Launch a model with advanced configuration")
+
     return parser
 
 
@@ -302,9 +312,9 @@ async def _get_frameworks(
     )
 
 
-async def _main(args: argparse.Namespace) -> None:
+async def _run_quickstart(args: argparse.Namespace) -> None:
     if not InitConfig.exists():
-        await _run_initial_configuration_wizard(args)
+        print("SML is not configured. Run `sml init` first.")
         return
 
     config = InitConfig.load()
@@ -345,6 +355,16 @@ async def _monitor() -> None:
     await LiveDisplay(state).run(_monitor())
 
 
+async def _main(args: argparse.Namespace) -> None:
+    subcommand = args.subcommand or ("quickstart" if InitConfig.exists() else "init")
+    if subcommand == "init":
+        await _run_initial_configuration_wizard(args)
+    elif subcommand == "quickstart":
+        await _run_quickstart(args)
+    elif subcommand == "advanced":
+        raise NotImplementedError("Advanced configuration is not yet implemented.")
+
+
 def main() -> None:
     args = _build_parser().parse_args()
     asyncio.run(_main(args))