feat: add composable deployment targets

catalog/components.yml

@@ -328,9 +328,9 @@ components:
   - {id: vercel, name: Vercel, layer: deploy, tier: platform, python: ">=3.11,<3.15"}
   - {id: railway, name: Railway, layer: deploy, tier: platform, python: ">=3.11,<3.15"}
   - {id: hf-spaces, name: Hugging Face Spaces, layer: deploy, tier: platform, python: ">=3.11,<3.15"}
-  - {id: modal, name: Modal, layer: deploy, tier: platform, python: ">=3.11,<3.15"}
-  - {id: runpod, name: RunPod, layer: deploy, tier: platform, python: ">=3.11,<3.15"}
-  - {id: bentocloud, name: BentoCloud, layer: deploy, tier: platform, python: ">=3.11,<3.15"}
+  - {id: modal, name: Modal, layer: deploy, tier: platform, python: ">=3.11,<3.15", packages: ["modal>=1.5.3,<2"]}
+  - {id: runpod, name: RunPod, layer: deploy, tier: platform, python: ">=3.11,<3.15", packages: ["runpod>=1.11.0,<2"]}
+  - {id: bentocloud, name: BentoCloud, layer: deploy, tier: platform, python: ">=3.11,<3.15", packages: ["bentoml>=1.4.39,<2"]}
   - {id: aws-ecs, name: AWS ECS, layer: deploy, tier: platform, python: ">=3.11,<3.15"}
   - {id: aws-sagemaker, name: AWS SageMaker, layer: deploy, tier: platform, python: ">=3.11,<3.15"}
   - {id: cloud-run, name: Google Cloud Run, layer: deploy, tier: platform, python: ">=3.11,<3.15"}

catalog/generated/catalog.json

@@ -1477,7 +1477,9 @@
       "python": ">=3.11,<3.15",
       "workloads": [],
       "roles": [],
-      "packages": []
+      "packages": [
+        "modal>=1.5.3,<2"
+      ]
     },
     {
       "id": "runpod",
@@ -1487,7 +1489,9 @@
       "python": ">=3.11,<3.15",
       "workloads": [],
       "roles": [],
-      "packages": []
+      "packages": [
+        "runpod>=1.11.0,<2"
+      ]
     },
     {
       "id": "bentocloud",
@@ -1497,7 +1501,9 @@
       "python": ">=3.11,<3.15",
       "workloads": [],
       "roles": [],
-      "packages": []
+      "packages": [
+        "bentoml>=1.4.39,<2"
+      ]
     },
     {
       "id": "aws-ecs",

copier.yml

@@ -496,6 +496,16 @@ deploy_target:
     Vertex AI: vertex-ai
     Azure Container Apps: azure-container-apps
     Azure ML: azure-ml
+  validator: >-
+    {% if workload == 'library' and deploy_target != 'none' %}
+    Libraries are published as packages; choose no deployment target
+    {% elif deploy_target in ['runpod', 'aws-sagemaker', 'vertex-ai', 'azure-ml'] and workload not in ['inference', 'hybrid'] %}
+    The selected ML platform requires an inference or hybrid workload
+    {% elif deploy_target in ['vercel', 'modal'] and workload in ['mcp', 'training'] %}
+    The selected serverless platform requires an HTTP-oriented workload
+    {% elif deploy_target == 'vercel' and workload == 'hybrid' %}
+    Vercel functions are not suitable for a hybrid training workspace
+    {% endif %}
 
 iac_provider:
   type: str
@@ -505,7 +515,7 @@ iac_provider:
     None: none
     Pulumi with Python: pulumi
     Terraform: terraform
-  when: "{{ deploy_target not in ['none', 'docker', 'vercel', 'hf-spaces'] }}"
+  when: "{{ deploy_target in ['aws-ecs', 'aws-sagemaker', 'cloud-run', 'vertex-ai', 'azure-container-apps', 'azure-ml'] }}"
 
 python_version:
   type: str
@@ -565,7 +575,7 @@ publish_to_pypi:
 use_docker:
   type: bool
   help: Include a multi-stage Dockerfile?
-  default: "{{ workload not in ['library', 'cli'] }}"
+  default: "{{ deploy_target != 'none' or workload not in ['library', 'cli'] }}"
 
 use_devcontainer:
   type: bool

pyproject.toml

@@ -22,6 +22,7 @@ package = false
 dev = [
     "copier>=9.17.1",
     "pydantic>=2.13.4,<3",
+    "python-hcl2>=8.1.2,<9",
     "pyyaml>=6.0.3,<7",
     "pytest>=9.1.1",
     "pytest-copie>=0.3.1",

scripts/compile_catalog.py

@@ -166,6 +166,7 @@ def build_artifacts(
             "  + ([sql_abstraction | default('none', true)] if sql_store not in ['none', 'supabase'] else [])",
             "  + [auth | default('none', true)]",
             "  + [serving | default('none', true)]",
+            "  + [deploy_target | default('none', true)]",
             "  + (training_extensions | default([]))",
             "  + (mlops_tools | default([]))",
             "  + (quality_tools | default([])) -%}",
@@ -187,6 +188,10 @@ def build_artifacts(
             "{%- if framework == 'pydantic-ai' and pydantic_ai_harness | default([]) %}",
             "    \"pydantic-ai-harness[{{ pydantic_ai_harness | join(',') }}]>=0.18.0,<0.19\",",
             "{%- endif %}",
+            "{%- if deploy_target | default('none', true) != 'none' and workload not in ['api', 'mcp', 'training'] and 'fastapi' not in interfaces | default([]) %}",
+            '    "fastapi[standard]>=0.141.1,<1",',
+            '    "uvicorn>=0.52.1,<1",',
+            "{%- endif %}",
             "",
         ]
     )

template/_catalog_dependencies.jinja

@@ -1,5 +1,5 @@
 {# Generated by scripts/compile_catalog.py. Do not edit. #}
-{%- set component_packages = {"accelerate": ["accelerate>=1.14.0,<2"], "ag2": ["ag2>=1.0.1,<2"], "agno": ["agno>=2.8.7,<3"], "anthropic": ["anthropic>=0.121.0,<1"], "autogen": ["autogen-agentchat>=0.7.5,<1"], "azure-openai": ["openai>=2.53.0,<3"], "beaver": ["beaver-db>=2.3.0,<3"], "bedrock": ["boto3>=1.43.67,<2"], "bentoml": ["bentoml>=1.4.39,<2"], "celery": ["celery>=5.6.3,<6"], "chainlit": ["chainlit>=2.11.1,<3"], "chroma": ["chromadb>=1.5.9,<2"], "cohere": ["cohere>=7.0.8,<8"], "crewai": ["crewai>=1.15.13,<2"], "dagster": ["dagster>=1.13.17,<2"], "datasets": ["datasets>=5.0.1,<6"], "deepeval": ["deepeval>=4.1.5,<5"], "deepseek": ["openai>=2.53.0,<3"], "dspy": ["dspy>=3.3.0,<4"], "duckdb": ["duckdb>=1.5.5,<2", "duckdb-engine>=0.17.0,<1"], "dvc": ["dvc>=3.67.1,<4"], "evidently": ["evidently>=0.7.21,<1"], "fastapi": ["fastapi[standard]>=0.141.1,<1", "pydantic>=2.13.4,<3", "uvicorn>=0.52.1,<1"], "fasthtml": ["python-fasthtml>=0.14.11,<0.15"], "fastmcp": ["fastmcp>=3.4.6,<4"], "feast": ["feast>=0.65.0,<1"], "fireworks": ["fireworks-ai>=1.2.6,<2"], "flask": ["flask>=3.1.2,<4"], "gemini": ["google-genai>=2.17.0,<3"], "google-adk": ["google-adk>=2.6.3,<3"], "gradio": ["gradio>=6.22.0,<7"], "groq": ["groq>=1.6.0,<2"], "haystack": ["haystack-ai>=3.0.0,<4"], "huggingface": ["huggingface-hub>=1.27.0,<2"], "jax": ["jax>=0.11.0,<0.12"], "jupyterlab": ["jupyterlab>=4.6.2,<5"], "keras": ["keras>=3.15.1,<4", "tensorflow>=2.21.0,<3"], "lancedb": ["lancedb>=0.36.0,<1"], "langchain": ["langchain>=1.3.14,<2"], "langfuse": ["langfuse>=4.14.3,<5"], "langgraph": ["langgraph>=1.2.10,<2"], "langsmith": ["langsmith>=0.10.17,<0.11"], "lightning": ["lightning>=2.6.5,<3"], "lingo": ["lingo-ai>=2.1.0,<3"], "litellm": ["litellm>=1.95.0,<2"], "llamaindex": ["llama-index>=0.14.23,<0.15"], "milvus": ["pymilvus>=3.0.1,<4"], "mistral": ["mistralai>=2.9.1,<3"], "mlflow": ["mlflow>=3.15.1,<4"], "mongodb": ["pymongo>=4.17.0,<5"], "mysql": ["pymysql>=1.2.0,<2"], "neo4j": ["neo4j>=6.2.0,<7"], "nicegui": ["nicegui>=3.15.0,<4"], "oidc": ["authlib>=1.7.2,<2"], "ollama": ["ollama>=0.6.2,<1"], "ollama-serving": ["ollama>=0.6.2,<1"], "openai": ["openai>=2.53.0,<3"], "openai-agents": ["openai-agents>=0.19.4,<1"], "openai-compatible": ["openai>=2.53.0,<3"], "openrouter": ["openrouter>=1.1.37,<2"], "opentelemetry": ["opentelemetry-sdk>=1.44.0,<2"], "optuna": ["optuna>=4.9.0,<5"], "pandera": ["pandera>=0.32.1,<0.33"], "peft": ["peft>=0.20.0,<0.21"], "pgvector": ["pgvector>=0.5.0,<1"], "phoenix": ["arize-phoenix>=19.19.1,<20"], "pinecone": ["pinecone>=9.1.0,<10"], "polars": ["polars>=1.43.2,<2"], "postgresql": ["psycopg[binary]>=3.3.4,<4"], "prefect": ["prefect>=3.8.2,<4"], "pydantic-ai": ["pydantic-ai>=2.27.0,<3"], "pydantic-evals": ["pydantic-evals>=2.27.0,<3"], "pytorch": ["torch>=2.13.0,<3"], "qdrant": ["qdrant-client>=1.19.0,<2"], "ragas": ["ragas>=0.4.3,<0.5"], "ray-serve": ["ray[serve]>=2.56.1,<3"], "redis": ["redis>=8.1.0,<9"], "scikit-learn": ["scikit-learn>=1.9.0,<2"], "sentence-transformers": ["sentence-transformers>=5.7.0,<6"], "smolagents": ["smolagents>=1.26.0,<2"], "sqlalchemy": ["sqlalchemy>=2.0.51,<3", "alembic>=1.19.1,<2"], "sqlmodel": ["sqlmodel>=0.0.39,<0.1", "sqlalchemy>=2.0.51,<3", "alembic>=1.19.1,<2"], "strands-agents": ["strands-agents>=1.51.0,<2"], "streamlit": ["streamlit>=1.61.1,<2"], "supabase": ["supabase>=2.31.0,<3"], "supabase-auth": ["supabase>=2.31.0,<3"], "temporal": ["temporalio>=1.31.0,<2"], "textual": ["textual>=8.2.8,<9"], "together": ["together>=2.30.0,<3"], "transformers": ["transformers>=5.14.1,<6", "torch>=2.13.0,<3"], "trl": ["trl>=1.9.2,<2"], "typer": ["typer>=0.27.1,<1"], "violetear": ["violetear[server]>=1.4.1,<2"], "vllm": ["vllm>=0.26.0,<0.27"], "wandb": ["wandb>=0.28.1,<0.29"], "weaviate": ["weaviate-client>=4.22.0,<5"], "xai": ["xai-sdk>=1.17.0,<2"], "xgboost": ["xgboost>=3.4.0,<4"]} -%}
+{%- set component_packages = {"accelerate": ["accelerate>=1.14.0,<2"], "ag2": ["ag2>=1.0.1,<2"], "agno": ["agno>=2.8.7,<3"], "anthropic": ["anthropic>=0.121.0,<1"], "autogen": ["autogen-agentchat>=0.7.5,<1"], "azure-openai": ["openai>=2.53.0,<3"], "beaver": ["beaver-db>=2.3.0,<3"], "bedrock": ["boto3>=1.43.67,<2"], "bentocloud": ["bentoml>=1.4.39,<2"], "bentoml": ["bentoml>=1.4.39,<2"], "celery": ["celery>=5.6.3,<6"], "chainlit": ["chainlit>=2.11.1,<3"], "chroma": ["chromadb>=1.5.9,<2"], "cohere": ["cohere>=7.0.8,<8"], "crewai": ["crewai>=1.15.13,<2"], "dagster": ["dagster>=1.13.17,<2"], "datasets": ["datasets>=5.0.1,<6"], "deepeval": ["deepeval>=4.1.5,<5"], "deepseek": ["openai>=2.53.0,<3"], "dspy": ["dspy>=3.3.0,<4"], "duckdb": ["duckdb>=1.5.5,<2", "duckdb-engine>=0.17.0,<1"], "dvc": ["dvc>=3.67.1,<4"], "evidently": ["evidently>=0.7.21,<1"], "fastapi": ["fastapi[standard]>=0.141.1,<1", "pydantic>=2.13.4,<3", "uvicorn>=0.52.1,<1"], "fasthtml": ["python-fasthtml>=0.14.11,<0.15"], "fastmcp": ["fastmcp>=3.4.6,<4"], "feast": ["feast>=0.65.0,<1"], "fireworks": ["fireworks-ai>=1.2.6,<2"], "flask": ["flask>=3.1.2,<4"], "gemini": ["google-genai>=2.17.0,<3"], "google-adk": ["google-adk>=2.6.3,<3"], "gradio": ["gradio>=6.22.0,<7"], "groq": ["groq>=1.6.0,<2"], "haystack": ["haystack-ai>=3.0.0,<4"], "huggingface": ["huggingface-hub>=1.27.0,<2"], "jax": ["jax>=0.11.0,<0.12"], "jupyterlab": ["jupyterlab>=4.6.2,<5"], "keras": ["keras>=3.15.1,<4", "tensorflow>=2.21.0,<3"], "lancedb": ["lancedb>=0.36.0,<1"], "langchain": ["langchain>=1.3.14,<2"], "langfuse": ["langfuse>=4.14.3,<5"], "langgraph": ["langgraph>=1.2.10,<2"], "langsmith": ["langsmith>=0.10.17,<0.11"], "lightning": ["lightning>=2.6.5,<3"], "lingo": ["lingo-ai>=2.1.0,<3"], "litellm": ["litellm>=1.95.0,<2"], "llamaindex": ["llama-index>=0.14.23,<0.15"], "milvus": ["pymilvus>=3.0.1,<4"], "mistral": ["mistralai>=2.9.1,<3"], "mlflow": ["mlflow>=3.15.1,<4"], "modal": ["modal>=1.5.3,<2"], "mongodb": ["pymongo>=4.17.0,<5"], "mysql": ["pymysql>=1.2.0,<2"], "neo4j": ["neo4j>=6.2.0,<7"], "nicegui": ["nicegui>=3.15.0,<4"], "oidc": ["authlib>=1.7.2,<2"], "ollama": ["ollama>=0.6.2,<1"], "ollama-serving": ["ollama>=0.6.2,<1"], "openai": ["openai>=2.53.0,<3"], "openai-agents": ["openai-agents>=0.19.4,<1"], "openai-compatible": ["openai>=2.53.0,<3"], "openrouter": ["openrouter>=1.1.37,<2"], "opentelemetry": ["opentelemetry-sdk>=1.44.0,<2"], "optuna": ["optuna>=4.9.0,<5"], "pandera": ["pandera>=0.32.1,<0.33"], "peft": ["peft>=0.20.0,<0.21"], "pgvector": ["pgvector>=0.5.0,<1"], "phoenix": ["arize-phoenix>=19.19.1,<20"], "pinecone": ["pinecone>=9.1.0,<10"], "polars": ["polars>=1.43.2,<2"], "postgresql": ["psycopg[binary]>=3.3.4,<4"], "prefect": ["prefect>=3.8.2,<4"], "pydantic-ai": ["pydantic-ai>=2.27.0,<3"], "pydantic-evals": ["pydantic-evals>=2.27.0,<3"], "pytorch": ["torch>=2.13.0,<3"], "qdrant": ["qdrant-client>=1.19.0,<2"], "ragas": ["ragas>=0.4.3,<0.5"], "ray-serve": ["ray[serve]>=2.56.1,<3"], "redis": ["redis>=8.1.0,<9"], "runpod": ["runpod>=1.11.0,<2"], "scikit-learn": ["scikit-learn>=1.9.0,<2"], "sentence-transformers": ["sentence-transformers>=5.7.0,<6"], "smolagents": ["smolagents>=1.26.0,<2"], "sqlalchemy": ["sqlalchemy>=2.0.51,<3", "alembic>=1.19.1,<2"], "sqlmodel": ["sqlmodel>=0.0.39,<0.1", "sqlalchemy>=2.0.51,<3", "alembic>=1.19.1,<2"], "strands-agents": ["strands-agents>=1.51.0,<2"], "streamlit": ["streamlit>=1.61.1,<2"], "supabase": ["supabase>=2.31.0,<3"], "supabase-auth": ["supabase>=2.31.0,<3"], "temporal": ["temporalio>=1.31.0,<2"], "textual": ["textual>=8.2.8,<9"], "together": ["together>=2.30.0,<3"], "transformers": ["transformers>=5.14.1,<6", "torch>=2.13.0,<3"], "trl": ["trl>=1.9.2,<2"], "typer": ["typer>=0.27.1,<1"], "violetear": ["violetear[server]>=1.4.1,<2"], "vllm": ["vllm>=0.26.0,<0.27"], "wandb": ["wandb>=0.28.1,<0.29"], "weaviate": ["weaviate-client>=4.22.0,<5"], "xai": ["xai-sdk>=1.17.0,<2"], "xgboost": ["xgboost>=3.4.0,<4"]} -%}
 {%- set selected_components = [framework]
   + (interfaces | default([]))
   + ([model_provider] if ai_capabilities in ['agents', 'rag', 'inference'] else [])
@@ -8,6 +8,7 @@
   + ([sql_abstraction | default('none', true)] if sql_store not in ['none', 'supabase'] else [])
   + [auth | default('none', true)]
   + [serving | default('none', true)]
+  + [deploy_target | default('none', true)]
   + (training_extensions | default([]))
   + (mlops_tools | default([]))
   + (quality_tools | default([])) -%}
@@ -29,3 +30,7 @@
 {%- if framework == 'pydantic-ai' and pydantic_ai_harness | default([]) %}
     "pydantic-ai-harness[{{ pydantic_ai_harness | join(',') }}]>=0.18.0,<0.19",
 {%- endif %}
+{%- if deploy_target | default('none', true) != 'none' and workload not in ['api', 'mcp', 'training'] and 'fastapi' not in interfaces | default([]) %}
+    "fastapi[standard]>=0.141.1,<1",
+    "uvicorn>=0.52.1,<1",
+{%- endif %}

template/readme.md.jinja

@@ -88,7 +88,7 @@ uv run poe --help             # list every task
 > reaches 1.0, drop `continue-on-error` from the `quality` job in
 > `.github/workflows/ci.yml`.
 
-{% if use_docker -%}
+{% if use_docker or deploy_target != 'none' -%}
 ## Docker
 
 ```bash

template/src/{{ module_name }}/{% if deploy_target != 'none' and workload not in ['api', 'mcp', 'training'] and 'fastapi' not in interfaces %}deployment.py{% endif %}.jinja

@@ -0,0 +1,44 @@
+"""Portable HTTP entrypoint for the selected deployment target."""
+
+from typing import Any
+
+from fastapi import FastAPI
+
+app = FastAPI(title="{{ project_name }}", version="0.1.0")
+
+
+@app.get("/health")
+async def health() -> dict[str, str]:
+    """Report deployment health without loading remote models."""
+    return {"status": "ok"}
+
+
+@app.get("/ping")
+async def ping() -> dict[str, str]:
+    """Implement the health contract used by managed ML endpoints."""
+    return {"status": "ok"}
+
+
+@app.post("/predict")
+@app.post("/invocations")
+async def predict(payload: dict[str, Any]) -> dict[str, Any]:
+    """Echo a validated envelope until project-specific inference is connected."""
+    return {"input": payload, "serving": "{{ serving }}"}
+
+
+@app.get("/")
+async def deployment_info() -> dict[str, str]:
+    """Describe the generated stack."""
+    return {
+        "project": "{{ project_slug }}",
+        "workload": "{{ workload }}",
+        "framework": "{{ framework }}",
+        "deployment": "{{ deploy_target }}",
+    }
+
+
+def main() -> None:
+    """Run the portable deployment entrypoint."""
+    import uvicorn
+
+    uvicorn.run(app, host="0.0.0.0", port=8000)

template/src/{{ module_name }}/{% if deploy_target == 'bentocloud' %}deploy{% endif %}/{% if deploy_target == 'bentocloud' %}bentoml_service.py{% endif %}.jinja

@@ -0,0 +1,12 @@
+"""BentoCloud service adapter."""
+
+import bentoml
+
+
+@bentoml.service
+class ProjectService:
+    """Minimal online service ready for model loading at startup."""
+
+    @bentoml.api
+    def health(self) -> dict[str, str]:
+        return {"status": "ok", "project": "{{ project_slug }}"}

template/src/{{ module_name }}/{% if deploy_target == 'modal' %}deploy{% endif %}/{% if deploy_target == 'modal' %}modal_app.py{% endif %}.jinja

@@ -0,0 +1,20 @@
+"""Modal deployment adapter."""
+
+import modal
+
+{% if workload == 'api' -%}
+from {{ module_name }}.api import app as web_app
+{%- elif 'fastapi' in interfaces | default([]) -%}
+from {{ module_name }}.interfaces.fastapi_app import app as web_app
+{%- else -%}
+from {{ module_name }}.deployment import app as web_app
+{%- endif %}
+
+app = modal.App("{{ project_slug }}")
+
+
+@app.function()
+@modal.asgi_app()
+def asgi_app():  # type: ignore[no-untyped-def]
+    """Expose the generated ASGI app through Modal."""
+    return web_app

template/src/{{ module_name }}/{% if deploy_target == 'runpod' %}deploy{% endif %}/{% if deploy_target == 'runpod' %}runpod_handler.py{% endif %}.jinja

@@ -0,0 +1,17 @@
+"""RunPod Serverless worker implementation."""
+
+from typing import Any
+
+import runpod
+
+
+def handler(job: dict[str, Any]) -> dict[str, Any]:
+    """Return a deterministic envelope ready for model inference code."""
+    return {
+        "input": job.get("input", {}),
+        "project": "{{ project_slug }}",
+        "serving": "{{ serving }}",
+    }
+
+
+RUNPOD_START = runpod.serverless.start

template/src/{{ module_name }}/{% if deploy_target in ['modal', 'runpod', 'bentocloud'] %}deploy{% endif %}/__init__.py.jinja

@@ -0,0 +1 @@
+"""Platform-specific deployment adapters."""

template/tests/{% if deploy_target != 'none' and workload not in ['api', 'mcp', 'training'] and 'fastapi' not in interfaces %}test_deployment.py{% endif %}.jinja

@@ -0,0 +1,13 @@
+from fastapi.testclient import TestClient
+
+from {{ module_name }}.deployment import app
+
+
+def test_deployment_health() -> None:
+    response = TestClient(app).get("/health")
+    assert response.status_code == 200
+    assert response.json() == {"status": "ok"}
+
+    prediction = TestClient(app).post("/predict", json={"value": 42})
+    assert prediction.status_code == 200
+    assert prediction.json()["input"] == {"value": 42}

template/{% if deploy_target != 'none' %}deploy{% endif %}/{% if deploy_target != 'none' %}README.md{% endif %}.jinja

@@ -0,0 +1,62 @@
+# Deployment: {{ deploy_target }}
+
+The checked-in `Dockerfile` is the portable deployment contract. Build and
+exercise it locally before publishing it to a registry:
+
+```bash
+docker build -t {{ project_slug }} .
+docker run --rm -p 8000:8000 {{ project_slug }}
+```
+
+Keep credentials out of images and configuration files. Configure the values
+documented in `.env.example` through the target platform's secret manager.
+
+{% if deploy_target == 'docker' -%}
+Start the local service with `docker compose up --build`.
+{%- elif deploy_target == 'render' -%}
+Create a Blueprint from this GitHub repository; Render discovers `render.yaml`
+and builds the root Dockerfile.
+{%- elif deploy_target == 'fly' -%}
+Run `fly launch --copy-config --no-deploy`, review `fly.toml`, set secrets with
+`fly secrets set`, then run `fly deploy`.
+{%- elif deploy_target == 'vercel' -%}
+Run `vercel link` once and `vercel deploy`; `api/index.py` exports the Python
+WSGI/ASGI application and `vercel.json` routes requests to it.
+{%- elif deploy_target == 'railway' -%}
+Create a service from the GitHub repository. Railway discovers `railway.toml`
+and the Dockerfile; configure variables in the service environment.
+{%- elif deploy_target == 'hf-spaces' -%}
+Push the repository to a Hugging Face Space. The uppercase `README.md` contains
+the required Docker Space metadata; development docs remain in `readme.md`.
+{%- elif deploy_target == 'modal' -%}
+Authenticate with `modal setup`, then run `modal deploy deploy/modal_app.py`.
+{%- elif deploy_target == 'runpod' -%}
+Build and push the image, create a Serverless endpoint from it, and set the
+container command to `python handler.py`.
+{%- elif deploy_target == 'bentocloud' -%}
+Run `bentoml build`, test the resulting Bento, then run `bentoml deploy`.
+{%- elif deploy_target == 'cloud-run' -%}
+Replace `IMAGE_URI` in `service.yaml`, then run
+`gcloud run services replace service.yaml`.
+{%- elif deploy_target == 'aws-ecs' -%}
+Push the image to ECR, replace placeholders in `deploy/task-definition.json`,
+register it with `aws ecs register-task-definition`, then update your service.
+{%- elif deploy_target == 'aws-sagemaker' -%}
+Push an inference image to ECR, replace placeholders in
+`deploy/sagemaker-model.json`, and create the model and endpoint resources.
+{%- elif deploy_target == 'vertex-ai' -%}
+Push the image to Artifact Registry, replace `IMAGE_URI` in
+`deploy/vertex-model.json`, upload the model, and deploy it to an endpoint.
+{%- elif deploy_target == 'azure-container-apps' -%}
+Push the image to ACR, replace placeholders in `deploy/container-app.yaml`, and
+run `az containerapp create --yaml deploy/container-app.yaml`.
+{%- elif deploy_target == 'azure-ml' -%}
+Replace the model/image placeholders, create `deploy/endpoint.yml`, then create
+`deploy/deployment.yml` with the Azure ML CLI v2.
+{%- endif %}
+
+{% if iac_provider != 'none' -%}
+The selected `{{ iac_provider }}` project lives under `infra/{{ iac_provider }}`.
+Its configuration intentionally requires registry images, roles, networks and
+regions as inputs instead of creating broad privileged infrastructure silently.
+{%- endif %}

template/{% if deploy_target == 'aws-ecs' %}deploy{% endif %}/{% if deploy_target == 'aws-ecs' %}task-definition.json{% endif %}.jinja

@@ -0,0 +1,16 @@
+{
+  "family": "{{ project_slug }}",
+  "networkMode": "awsvpc",
+  "requiresCompatibilities": ["FARGATE"],
+  "cpu": "512",
+  "memory": "1024",
+  "executionRoleArn": "EXECUTION_ROLE_ARN",
+  "containerDefinitions": [
+    {
+      "name": "app",
+      "image": "IMAGE_URI",
+      "essential": true,
+      "portMappings": [{"containerPort": 8000, "protocol": "tcp"}]
+    }
+  ]
+}