fix: update deployment prompts and messages (#129)

* fix: update deployment prompts and messages

* fix: fix prompt list and expand by default

* chore: clean up

Author: yanbasic

src/emd/commands/deploy.py

@@ -28,6 +28,8 @@
 from emd.utils.decorators import catch_aws_credential_errors,check_emd_env_exist,load_aws_profile
 from emd.utils.logger_utils import make_layout
 from emd.utils.exceptions import ModelNotSupported,ServiceNotSupported,InstanceNotSupported
+from prompt_toolkit import prompt
+from prompt_toolkit.completion import FuzzyWordCompleter
 
 app = typer.Typer(pretty_exceptions_enable=False)
 console = Console()
@@ -52,13 +54,26 @@ def show_help(choice):
     return f"{choice} (shortcut)"
 
 
-def supported_models_filter(region:str,support_models:list[Model]):
+def supported_models_filter(
+    region:str,
+    allow_local_deploy,
+    only_allow_local_deploy,
+    support_models:list[Model]
+):
     ret = []
     is_cn_region = check_cn_region(region)
 
     for model in support_models:
         if is_cn_region and not model.allow_china_region:
             continue
+
+        # Skip models that only support local services when local deployment is not allowed
+        if not allow_local_deploy:
+            # Check if all supported services are local services
+            all_local_services = all(service.service_type == ServiceType.LOCAL for service in model.supported_services)
+            if all_local_services:
+                continue
+
         ret.append(model)
     return ret
 
@@ -130,72 +145,60 @@ def is_valid_model_tag(name,pattern=MODEL_TAG_PATTERN):
     return bool(re.match(pattern, name))
 
 
-def ask_model_id(region,model_id=None):
+# Define a natural sort key function to handle numeric values in model names
+def natural_sort_key(s):
+    # Split the string into text and numeric parts
+    return [int(c) if c.isdigit() else float(c) if c.replace('.', '', 1).isdigit() else c.lower()
+            for c in re.split(r'(\d+\.\d+|\d+)', s)]
+
+
+def ask_model_id(region, allow_local_deploy, only_allow_local_deploy, model_id=None):
     if model_id is not None:
         return model_id
 
-    # step 1: select model series name
-    support_models:list[Model] = sorted(
-        [Model.get_model(m) for m in Model.get_supported_models()
-         if hasattr(Model.get_model(m), 'model_series') and hasattr(Model.get_model(m).model_series, 'model_series_name')],
-        key=lambda x:x.model_series.model_series_name
-    )
-    # filter models
-    support_models = supported_models_filter(region,support_models)
+    try:
+        supported_models = [Model.get_model(m) for m in Model.get_supported_models()]
+        filtered_models = supported_models_filter(region, allow_local_deploy, only_allow_local_deploy, supported_models)
 
-    if not support_models:
-        raise ModelNotSupported(region)
+        if not filtered_models:
+            raise ModelNotSupported(region)
 
-    model_series_map = defaultdict(list)
-    for model in support_models:
-        model_series_map[model.model_series.model_series_name].append(model)
-
-    def _get_series_description(models:list[Model]):
-        model = models[0]
-        description = "\n"
-        description += model.model_series.description
-        description += f"\nreference link: {model.model_series.reference_link}"
-        description += "\nSupported models: "+ "\n - " + "\n - ".join(model.model_id for model in models)
-        return description
-
-    series_name = select_with_help(
-        "Select the model series:",
-        choices=[
-            Choice(
-                title=series_name,
-                description=_get_series_description(models),
-            )
-            for series_name,models in model_series_map.items()
-        ],
-        show_description=True,
-        style=custom_style
-    ).ask()
-    if series_name is None:
-        raise typer.Exit(0)
+        model_ids = sorted([model.model_id for model in filtered_models], key=natural_sort_key)
+        completer = FuzzyWordCompleter(model_ids, WORD=True)
 
-    def _get_model_description(model:Model):
-        description=f"\n\nModelType: {model.model_type}\nApplication Scenario: {model.application_scenario}"
-        if model.description:
-            description += f"\nDescription: {model.description}"
-        return description
-
-    # step 2 select model_id
-    model_id = select_with_help(
-        "Select the model name:",
-        choices=[
-            Choice(
-                title=model.model_id,
-                description=_get_model_description(model)
-            )
-            for model in model_series_map[series_name]
-        ],
-        show_description=True,
-        style=custom_style
-    ).ask()
-
-    if model_id is None:
-        raise typer.Exit(0)
-    return model_id
+        from prompt_toolkit.formatted_text import HTML
+        from prompt_toolkit import PromptSession
+        from prompt_toolkit.application.current import get_app
+
+        session = PromptSession(
+            completer=completer,
+            complete_while_typing=True,
+        )
+
+        def get_prompt_message():
+            buffer = get_app().current_buffer
+            if buffer.text:
+                return HTML('<b>? Enter model name: </b>')
+            else:
+                return HTML('<b>? Enter model name: </b><span fg="#888888">(Type to search, run "emd list-supported-models" for full model list)</span>')
+
+        selected_model = session.prompt(get_prompt_message, pre_run=lambda: get_app().current_buffer.start_completion())
+
+        if not selected_model:
+            console.print("[bold yellow]Model selection cancelled[/bold yellow]")
+            raise typer.Exit(0)
+
+        if selected_model not in model_ids:
+            console.print(f"[bold red]Invalid model name: {selected_model}[/bold red]")
+            raise typer.Exit(1)
+
+        return selected_model
+
+    except Exception as e:
+        if not isinstance(e, (ModelNotSupported, typer.Exit)):
+            console.print(f"[bold red]Error during model selection: {str(e)}[/bold red]")
+            raise typer.Exit(1)
+        raise
 
 
 #@app.callback(invoke_without_command=True)(invoke_without_command=True)
@@ -268,7 +271,12 @@ def deploy(
 
     vpc_id = None
     # ask model id
-    model_id = ask_model_id(region,model_id=model_id)
+    model_id = ask_model_id(
+        region,
+        allow_local_deploy,
+        only_allow_local_deploy,
+        model_id=model_id
+    )
 
     if not check_model_support_on_cn_region(model_id,region):
         raise ModelNotSupported(region,model_id=model_id)
@@ -286,7 +294,7 @@ def deploy(
     if service_type is None:
         if len(supported_services) > 1:
             service_name = select_with_help(
-                "Select the service for deployment:",
+                "Select model hosting service:",
                 choices=[
                     Choice(
                         title=service.name,
@@ -328,7 +336,7 @@ def deploy(
                 vpc_name = next((tag['Value'] for tag in vpc.get('Tags', []) if tag.get('Key') == 'Name'), None)
                 vpc['Name'] = vpc_name if vpc_name else '-'
             emd_vpc = select_with_help(
-                "Select the VPC (Virtual Private Cloud) you want to deploy the ESC service:",
+                "Select VPC (Virtual Private Cloud):",
                 choices=[
                     Choice(
                         title=f"{emd_default_vpc['VpcId']} ({emd_default_vpc['CidrBlock']}) (EMD-vpc)" if emd_default_vpc else 'Create a new VPC',
@@ -412,7 +420,7 @@ def deploy(
         if instance_type is None:
             if len(supported_instances)>1:
                 instance_type = select_with_help(
-                    "Select the instance type:",
+                    "Select instance type:",
                     choices=[
                         Choice(
                             title=instance.instance_type,
@@ -449,7 +457,7 @@ def deploy(
     if engine_type is None:
         if len(supported_engines)>1:
             engine_type = select_with_help(
-                "Select the inference engine to use:",
+                "Select inference engine:",
                 choices=[
                     Choice(
                         title=engine.engine_type,
@@ -476,7 +484,7 @@ def deploy(
     if framework_type is None:
         if len(supported_frameworks)>1:
             framework_type = select_with_help(
-                "Select the inference engine to use:",
+                "Select inference engine:",
                 choices=[
                     Choice(
                         title=framework.framework_type,
@@ -488,7 +496,6 @@ def deploy(
             ).ask()
         else:
             framework_type = supported_frameworks[0].framework_type
-            console.print(f"[bold blue]framework type: {framework_type}[/bold blue]")
     else:
         supported_framework_types = model.supported_framework_types
         console.print(f"[bold blue]framework type: {framework_type}[/bold blue]")
@@ -502,8 +509,8 @@ def deploy(
     if extra_params is None:
         while True:
             extra_params = questionary.text(
-                "(Optional) Additional deployment parameters (JSON string or local file path), you can skip by pressing Enter:",
-                instruction="Parameters format: https://aws-samples.github.io/easy-model-deployer/en/installation/#-extra-params.",
+                "(Optional) Additional parameters, you can skip by pressing Enter:",
+                instruction="Usage: https://aws-samples.github.io/easy-model-deployer/en/best_deployment_practices/#extra-parameters-usage",
                 default="{}"
             ).ask()
 
@@ -528,7 +535,7 @@ def deploy(
     if not skip_confirm and not service_type == ServiceType.LOCAL:
         while True:
             model_tag = questionary.text(
-                    "(Optional) Add a model deployment tag (custom label), you can skip by pressing Enter:",
+                    "(Optional) Custom tag (label), you can skip by pressing Enter:",
                     default=MODEL_DEFAULT_TAG
                 ).ask()
             # if model_tag == MODEL_DEFAULT_TAG:
@@ -547,7 +554,7 @@ def deploy(
 
     if not skip_confirm:
         if not typer.confirm(
-            "Would you like to proceed with the deployment? Please verify your selections above.",
+            "Ready to deploy? Please confirm your selections above.",
             abort=True,
         ):
             raise typer.Exit(0)

src/emd/models/services.py

@@ -21,7 +21,7 @@
         "SageMakerEndpointName": ValueWithDefault(name="sagemaker_endpoint_name",default="Auto-generate"),
         "APIKey": ValueWithDefault(name="api_key",default="")
     },
-    name = "Amazon SageMaker AI Real-time inference with OpenAI Compatible API",
+    name = "Amazon SageMaker AI Real-time inference with OpenAI-Compatible API",
     service_type=ServiceType.SAGEMAKER,
     description="Amazon SageMaker Real-time inference provides low-latency, interactive inference through fully managed endpoints that support autoscaling. It provides an OpenAI-compatible REST API (e.g., /v1/completions) via an Application Load Balancer (ALB).\n(https://docs.aws.amazon.com/sagemaker/latest/dg/realtime-endpoints.html)",
     support_cn_region = True
@@ -64,7 +64,7 @@
         "AutoScalingTargetValue": ValueWithDefault(name="auto_scaling_target_value",default=10),
         "APIKey": ValueWithDefault(name="api_key",default="")
     },
-    name = "Amazon SageMaker AI Asynchronous inference with OpenAI Compatible API",
+    name = "Amazon SageMaker AI Asynchronous inference with OpenAI-Compatible API",
     service_type=ServiceType.SAGEMAKER_ASYNC,
     description="Amazon SageMaker Asynchronous Inference queues requests for processing asynchronously, making it suitable for large payloads (up to 1GB) and long processing times (up to one hour), while also enabling cost savings by autoscaling to zero when idle. It provides an OpenAI-compatible REST API (e.g., /v1/completions) via an Application Load Balancer (ALB).\n(https://docs.aws.amazon.com/sagemaker/latest/dg/async-inference.html)",
     support_cn_region = True
@@ -107,7 +107,7 @@
         "ContainerGpu":"instance_gpu_num",
         "APIKey": ValueWithDefault(name="api_key",default="")
     },
-    name = "Amazon ECS with OpenAI Compatible API",
+    name = "Amazon ECS with OpenAI-Compatible API",
     service_type=ServiceType.ECS,
     description="Amazon Elastic Container Service is a fully managed service that runs containerized applications in clusters with auto scaling. It provides an OpenAI-compatible REST API (e.g., /v1/completions) via an Application Load Balancer (ALB), enabling integration with AI models for tasks like chatbots or document analysis. (https://docs.aws.amazon.com/AmazonECS/latest/developerguide)",
     support_cn_region = True,