[Feature][Executor] Refine the image startup scripts to support customized executor image

src/promptflow-core/promptflow/execution_app.py

@@ -0,0 +1,116 @@
+import multiprocessing
+import os
+import subprocess
+import sys
+from pathlib import Path
+
+from promptflow._utils.logger_utils import service_logger
+
+
+class RunMode:
+    COMPUTE = "compute"
+    SERVING = "serving"
+
+
+def get_process_num() -> int:
+    """Get process number of server based on CPU core num."""
+
+    worker_num_env = "PROMPTFLOW_WORKER_NUM"
+    worker_num = os.getenv(worker_num_env)
+    cpu_core_num = multiprocessing.cpu_count()
+    try:
+        if worker_num is None:
+            service_logger.info(
+                f"The promptflow worker number not set, setting {worker_num_env} to CPU core number: {cpu_core_num}"
+            )
+            return cpu_core_num
+        else:
+            worker_num = int(worker_num)
+            max_process = cpu_core_num * 2
+            if worker_num > max_process:
+                service_logger.info(
+                    "The promptflow worker number is too large, "
+                    f"setting {worker_num_env} to (2 * the number of cores) = {max_process}"
+                )
+                return max_process
+            else:
+                return worker_num
+    except Exception as e:
+        service_logger.error(f"Invalid value for {worker_num_env}. It must be an integer. Error: {e}")
+        return cpu_core_num
+
+
+def is_pf_model(model_dir: Path) -> bool:
+    """
+    Check whether a given model_dir contains 'MLmodel' file or 'flow.dag.yaml'.
+    """
+    return model_dir.joinpath("MLmodel").exists() or model_dir.joinpath("flow.dag.yaml").exists()
+
+
+def get_run_mode() -> str:
+    """Get PROMPTFLOW_RUN_MODE for pf serving scenario."""
+
+    run_mode = os.getenv("PROMPTFLOW_RUN_MODE")
+    if not run_mode:
+        service_logger.info("Detecting promptflow run mode...")
+        model_dir = os.getenv("AZUREML_MODEL_DIR")
+        pf_project_path = os.getenv("PROMPTFLOW_PROJECT_PATH")
+        if not model_dir and not pf_project_path:
+            run_mode = RunMode.COMPUTE
+        elif model_dir:
+            # Check model file to determine if it is MIR serving deployment or runtime deployment.
+            # will remove this once we didn't support MIR runtime for both 1P & 3P.
+            model_dir = Path(model_dir)
+            if is_pf_model(model_dir):
+                run_mode = RunMode.SERVING
+            elif len(list(model_dir.iterdir())) == 1 and is_pf_model(list(model_dir.iterdir())[0]):
+                # Only has one sub dir, it is a model deployment
+                run_mode = RunMode.SERVING
+            else:
+                run_mode = RunMode.COMPUTE
+        else:
+            # If PROMPTFLOW_PROJECT_PATH is set, this is a serving environment
+            run_mode = RunMode.SERVING
+
+    service_logger.info(f"Promptflow run mode: {run_mode}")
+    return run_mode
+
+
+def start_server():
+    """Start promptflow server based on run mode."""
+
+    run_mode = get_run_mode()
+    if run_mode == RunMode.SERVING:
+        worker_num = str(get_process_num())
+        worker_threads = os.getenv("PROMPTFLOW_WORKER_THREADS", "1")
+        gunicorn_app = "promptflow.core._serving.app:create_app(extension_type='azureml')"
+        service_logger.info(
+            f"Start promptflow serving server with worker_num: {worker_num}, "
+            f"worker_threads: {worker_threads}, app: {gunicorn_app}"
+        )
+        subprocess.run(
+            [
+                "gunicorn",
+                "-w",
+                worker_num,
+                "--threads",
+                worker_threads,
+                "-b",
+                "0.0.0.0:8080",
+                "--timeout",
+                "300",
+                gunicorn_app,
+            ],
+            stdout=sys.stdout,
+            stderr=sys.stderr,
+        )
+    else:
+        host = os.getenv("HOST", "0.0.0.0")
+        port = os.getenv("PORT", "8000")
+        uvicorn_app = "promptflow.executor._service.app:app"
+        service_logger.info(f"Start promptflow python server with app: {uvicorn_app}")
+        subprocess.run(
+            ["gunicorn", "-k", "uvicorn.workers.UvicornWorker", "-b", f"{host}:{port}", uvicorn_app],
+            stdout=sys.stdout,
+            stderr=sys.stderr,
+        )

src/promptflow-core/pyproject.toml

@@ -51,16 +51,19 @@ flask = ">=2.2.3,<4.0.0"  # Serving endpoint requirements
 python-dateutil = ">=2.1.0,<3.0.0"  # Contracts RunInfo
 # ========executor-service dependencies========
 fastapi = ">=0.109.0,<1.0.0" # used to build web executor server
+uvicorn = ">=0.27.0,<1.0.0" # used to run web executor server
 # ========azureml-serving dependencies========
 # AzureML connection dependencies
 azure-identity = ">=1.12.0,<2.0.0"
 azure-ai-ml = ">=1.14.0,<2.0.0"
 # MDC dependencies for monitoring
 azureml-ai-monitoring = ">=0.1.0b3,<1.0.0"
+# Serving app dependencies
+gunicorn = ">=21.2.0,<22.0.0" # used to run serving app server
 
 [tool.poetry.extras]
-executor-service = ["fastapi"]
-azureml-serving = ["azure-identity", "azure-ai-ml", "azureml-ai-monitoring"]
+executor-service = ["fastapi", "uvicorn"]
+azureml-serving = ["azure-identity", "azure-ai-ml", "azureml-ai-monitoring", "gunicorn"]
 
 [tool.poetry.group.dev.dependencies]
 pre-commit = "*"
@@ -77,6 +80,9 @@ requires = [
 ]
 build-backend = "poetry.core.masonry.api"
 
+[tool.poetry.scripts]
+pfapp = "promptflow.execution_app:start_server"
+
 [tool.pytest.ini_options]
 markers = [
   "unittest",