flowmesh-cli-stack 0.1.9__tar.gz → 0.1.10rc2__tar.gz

{flowmesh_cli_stack-0.1.9/src/flowmesh_cli_stack.egg-info → flowmesh_cli_stack-0.1.10rc2}/PKG-INFO

@@ -1,13 +1,13 @@
 Metadata-Version: 2.4
 Name: flowmesh-cli-stack
-Version: 0.1.9
+Version: 0.1.10rc2
 Summary: FlowMesh CLI stack commands
 License-Expression: Apache-2.0
 Requires-Python: >=3.12
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: flowmesh-cli==0.1.9
-Requires-Dist: flowmesh-sdk-stack==0.1.9
+Requires-Dist: flowmesh-cli==0.1.10rc2
+Requires-Dist: flowmesh-sdk-stack==0.1.10rc2
 Requires-Dist: docker>=7.1.0
 Requires-Dist: packaging>=24
 Requires-Dist: pyyaml>=6.0.2

{flowmesh_cli_stack-0.1.9 → flowmesh_cli_stack-0.1.10rc2}/pyproject.toml

@@ -4,15 +4,15 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "flowmesh-cli-stack"
-version = "0.1.9"
+version = "0.1.10rc2"
 description = "FlowMesh CLI stack commands"
 readme = "README.md"
 requires-python = ">=3.12"
 license = "Apache-2.0"
 license-files = ["LICENSE"]
 dependencies = [
-    "flowmesh-cli==0.1.9",
-    "flowmesh-sdk-stack==0.1.9",
+    "flowmesh-cli==0.1.10rc2",
+    "flowmesh-sdk-stack==0.1.10rc2",
     "docker>=7.1.0",
     "packaging>=24",
     "pyyaml>=6.0.2",

{flowmesh_cli_stack-0.1.9 → flowmesh_cli_stack-0.1.10rc2}/src/flowmesh_cli_stack/assets/.env.example

@@ -181,6 +181,14 @@ SERVER_CUDA_PROBE_IMAGE=nvidia/cuda:12.9.1-base-ubuntu24.04
 # Optional Docker runtime name for GPU containers.
 DOCKER_GPU_RUNTIME=nvidia
 CUDA_VISIBLE_DEVICES=all
+# Report a GPU as unavailable while a process outside FlowMesh uses it.
+WORKER_FOREIGN_GPU_GATE=true
+# Foreign GPU-memory threshold in MiB.
+WORKER_FOREIGN_GPU_MEM_MIB=1024
+# Consecutive occupancy readings required before a device changes availability.
+WORKER_FOREIGN_GPU_CONSECUTIVE=2
+# Delay occupancy checks after a worker task completes.
+WORKER_FOREIGN_GPU_GRACE_SEC=90
 WORKER_UPLOAD_RESULTS=false
 WORKER_EXECUTOR_IDLE_CLEANUP_SEC=60
 MODEL_CLEANUP_AFTER_UPLOAD=0

{flowmesh_cli_stack-0.1.9 → flowmesh_cli_stack-0.1.10rc2}/src/flowmesh_cli_stack/env_schema.py

@@ -593,6 +593,39 @@ STACK_ENV_SCHEMA = EnvSchema(
                 EnvVar(
                     "CUDA_VISIBLE_DEVICES", "all", var_type=EnvVarType.CSV_INTS_OR_ALL
                 ),
+                EnvVar(
+                    "WORKER_FOREIGN_GPU_GATE",
+                    "true",
+                    description=[
+                        "Report a GPU as unavailable while a process outside "
+                        "FlowMesh uses it.",
+                    ],
+                    var_type=EnvVarType.BOOL,
+                ),
+                EnvVar(
+                    "WORKER_FOREIGN_GPU_MEM_MIB",
+                    "1024",
+                    description="Foreign GPU-memory threshold in MiB.",
+                    var_type=EnvVarType.INT,
+                    min_value=1,
+                ),
+                EnvVar(
+                    "WORKER_FOREIGN_GPU_CONSECUTIVE",
+                    "2",
+                    description=[
+                        "Consecutive occupancy readings required before a device "
+                        "changes availability.",
+                    ],
+                    var_type=EnvVarType.INT,
+                    min_value=1,
+                ),
+                EnvVar(
+                    "WORKER_FOREIGN_GPU_GRACE_SEC",
+                    "90",
+                    description="Delay occupancy checks after a worker task completes.",
+                    var_type=EnvVarType.FLOAT,
+                    min_value=0,
+                ),
                 EnvVar("WORKER_UPLOAD_RESULTS", "false", var_type=EnvVarType.BOOL),
                 EnvVar(
                     "WORKER_EXECUTOR_IDLE_CLEANUP_SEC",

{flowmesh_cli_stack-0.1.9 → flowmesh_cli_stack-0.1.10rc2}/src/flowmesh_cli_stack/worker.py

@@ -80,11 +80,11 @@ def worker_up(
             "Repeat --config-raw to provide multiple configs."
         ),
     ),
-    name_template: str | None = typer.Option(
+    alias_template: str | None = typer.Option(
         None,
-        "--name-template",
+        "--alias-template",
         help=(
-            "Worker name template for cpu/gpu presets. Placeholders: "
+            "Worker alias template for cpu/gpu presets. Placeholders: "
             "{slug}, {kind}, {idx}, {gpu}. "
             "Default: '{slug}_worker_{kind}_{idx|gpu}'."
         ),
@@ -113,21 +113,21 @@ def worker_up(
             targets=targets,
             config_paths=config,
             config_raw=config_raw,
-            name_template=name_template,
+            alias_template=alias_template,
             slug=slug,
         )
     except FlowMeshError as exc:
         logging.error(str(exc))
         raise typer.Exit(code=1)
     for label, worker_info in created:
-        worker_name = worker_info.get("name", "<unknown>")
-        logging.success(f"Created {label} '{worker_name}':")
+        worker_alias = worker_info.get("alias", "<unknown>")
+        logging.success(f"Created {label} '{worker_alias}':")
         logging.log(json.dumps(worker_info, indent=2))
 
 
 @app.command("start")
 def worker_start(
-    names: list[str] = typer.Argument(..., help="Worker name(s) or 'all'"),
+    aliases: list[str] = typer.Argument(..., help="Worker alias(es) or 'all'"),
     env_file: Path = typer.Option(
         DEFAULT_ENV_FILE, "--env-file", help="Env file to load defaults"
     ),
@@ -139,20 +139,20 @@ def worker_start(
     """Start a stopped worker container."""
     client = stack_node_client(env_file, base_url, token or None)
     try:
-        started = operate_workers(client, names, operation="start")
+        started = operate_workers(client, aliases, operation="start")
     except FlowMeshError as exc:
         logging.error(str(exc))
         raise typer.Exit(code=1)
     if not started:
         logging.warning("No workers found.")
         return
-    for name in started:
-        logging.success(f"Started worker {name}")
+    for alias in started:
+        logging.success(f"Started worker {alias}")
 
 
 @app.command("stop")
 def worker_stop(
-    names: list[str] = typer.Argument(..., help="Worker name(s) or 'all'"),
+    aliases: list[str] = typer.Argument(..., help="Worker alias(es) or 'all'"),
     env_file: Path = typer.Option(
         DEFAULT_ENV_FILE, "--env-file", help="Env file to load defaults"
     ),
@@ -164,20 +164,20 @@ def worker_stop(
     """Stop a running worker container without removing it."""
     client = stack_node_client(env_file, base_url, token or None)
     try:
-        stopped = operate_workers(client, names, operation="stop")
+        stopped = operate_workers(client, aliases, operation="stop")
     except FlowMeshError as exc:
         logging.error(str(exc))
         raise typer.Exit(code=1)
     if not stopped:
         logging.warning("No workers found.")
         return
-    for name in stopped:
-        logging.success(f"Stopped worker {name}")
+    for alias in stopped:
+        logging.success(f"Stopped worker {alias}")
 
 
 @app.command("down")
 def worker_down(
-    names: list[str] = typer.Argument(..., help="Worker name(s) or 'all'"),
+    aliases: list[str] = typer.Argument(..., help="Worker alias(es) or 'all'"),
     env_file: Path = typer.Option(
         DEFAULT_ENV_FILE, "--env-file", help="Env file to load defaults"
     ),
@@ -188,9 +188,9 @@ def worker_down(
 ) -> None:
     """Destroy a worker or all workers, removing containers and associated resources."""
     client = stack_node_client(env_file, base_url, token or None)
-    if "all" in names:
-        if len(names) != 1:
-            logging.error("Use either 'all' or worker names, not both.")
+    if "all" in aliases:
+        if len(aliases) != 1:
+            logging.error("Use either 'all' or worker aliases, not both.")
             raise typer.Exit(code=1)
         try:
             logging.info("Destroying all workers...")
@@ -202,15 +202,15 @@ def worker_down(
         return
 
     try:
-        destroyed = operate_workers(client, names, operation="destroy")
+        destroyed = operate_workers(client, aliases, operation="destroy")
     except FlowMeshError as exc:
         logging.error(str(exc))
         raise typer.Exit(code=1)
     if not destroyed:
         logging.warning("No workers found.")
         return
-    for name in destroyed:
-        logging.success(f"Destroyed worker {name}")
+    for alias in destroyed:
+        logging.success(f"Destroyed worker {alias}")
 
 
 @app.command("pull")

{flowmesh_cli_stack-0.1.9 → flowmesh_cli_stack-0.1.10rc2/src/flowmesh_cli_stack.egg-info}/PKG-INFO

@@ -1,13 +1,13 @@
 Metadata-Version: 2.4
 Name: flowmesh-cli-stack
-Version: 0.1.9
+Version: 0.1.10rc2
 Summary: FlowMesh CLI stack commands
 License-Expression: Apache-2.0
 Requires-Python: >=3.12
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: flowmesh-cli==0.1.9
-Requires-Dist: flowmesh-sdk-stack==0.1.9
+Requires-Dist: flowmesh-cli==0.1.10rc2
+Requires-Dist: flowmesh-sdk-stack==0.1.10rc2
 Requires-Dist: docker>=7.1.0
 Requires-Dist: packaging>=24
 Requires-Dist: pyyaml>=6.0.2

flowmesh_cli_stack-0.1.10rc2/src/flowmesh_cli_stack.egg-info/requires.txt

@@ -0,0 +1,5 @@
+flowmesh-cli==0.1.10rc2
+flowmesh-sdk-stack==0.1.10rc2
+docker>=7.1.0
+packaging>=24
+pyyaml>=6.0.2

flowmesh_cli_stack-0.1.9/src/flowmesh_cli_stack.egg-info/requires.txt

@@ -1,5 +0,0 @@
-flowmesh-cli==0.1.9
-flowmesh-sdk-stack==0.1.9
-docker>=7.1.0
-packaging>=24
-pyyaml>=6.0.2