feat: remote development (#4942)

* feat: remote development

Signed-off-by: Frost Ming <me@frostming.com>

* feat: tail pod logs

Signed-off-by: Frost Ming <me@frostming.com>

* fix: change the color order

Signed-off-by: Frost Ming <me@frostming.com>

* fix: add cluster name

Signed-off-by: Frost Ming <me@frostming.com>

* feat: connect to existing dev deployment

Signed-off-by: Frost Ming <me@frostming.com>

* feat: develop command

Signed-off-by: Frost Ming <me@frostming.com>

* fix: reload service when importing

Signed-off-by: Frost Ming <me@frostming.com>

* fix: upload models for bare push bento

Signed-off-by: Frost Ming <me@frostming.com>

---------

Signed-off-by: Frost Ming <me@frostming.com>

src/_bentoml_impl/loader.py

@@ -163,7 +163,6 @@ def import_service(
         assert (
             module_name and attrs_str
         ), f'Invalid import target "{service_identifier}", must format as "<module>:<attribute>"'
-
         module = importlib.import_module(module_name)
         root_service_name, _, depend_path = attrs_str.partition(".")
         root_service = t.cast("Service[t.Any]", getattr(module, root_service_name))

src/_bentoml_impl/server/serving.py

@@ -332,7 +332,6 @@ def serve_http(
         if development_mode:
             arbiter_kwargs["debug"] = True
             arbiter_kwargs["loggerconfig"] = SERVER_LOGGING_CONFIG
-            arbiter_kwargs["loglevel"] = "WARNING"
 
         arbiter = create_standalone_arbiter(**arbiter_kwargs)
         arbiter.exit_stack.enter_context(

src/bentoml/_internal/bento/bento.py

@@ -53,6 +53,7 @@
     from _bentoml_sdk import Service as NewService
     from _bentoml_sdk.service import ServiceConfig
 
+    from ..cloud.schemas.modelschemas import BentoManifestSchema
     from ..models import Model as StoredModel
     from ..service import Service
     from ..service.inference_api import InferenceAPI
@@ -176,6 +177,49 @@ def _fs(self) -> FS:
     def info(self) -> BentoInfo:
         return self._info
 
+    @property
+    def entry_service(self) -> str:
+        return self.info.entry_service
+
+    def get_manifest(self) -> BentoManifestSchema:
+        from ..cloud.schemas.modelschemas import BentoManifestSchema
+        from ..cloud.schemas.modelschemas import BentoRunnerResourceSchema
+        from ..cloud.schemas.modelschemas import BentoRunnerSchema
+
+        info = self.info
+        models = [str(m.tag) for m in info.all_models]
+        runners = [
+            BentoRunnerSchema(
+                name=r.name,
+                runnable_type=r.runnable_type,
+                models=r.models,
+                resource_config=(
+                    BentoRunnerResourceSchema(
+                        cpu=r.resource_config.get("cpu"),
+                        nvidia_gpu=r.resource_config.get("nvidia.com/gpu"),
+                        custom_resources=r.resource_config.get("custom_resources"),
+                    )
+                    if r.resource_config
+                    else None
+                ),
+            )
+            for r in info.runners
+        ]
+        return BentoManifestSchema(
+            name=info.name,
+            entry_service=info.entry_service,
+            service=info.service,
+            bentoml_version=info.bentoml_version,
+            apis={},
+            models=models,
+            runners=runners,
+            size_bytes=self.total_size(),
+            services=info.services,
+            envs=info.envs,
+            schema=info.schema,
+            version=info.version,
+        )
+
     @classmethod
     @inject
     def create(
@@ -184,6 +228,8 @@ def create(
         version: t.Optional[str] = None,
         build_ctx: t.Optional[str] = None,
         platform: t.Optional[str] = None,
+        bare: bool = False,
+        reload: bool = False,
     ) -> Bento:
         from _bentoml_sdk.models import BentoModel
 
@@ -203,7 +249,10 @@ def create(
         BentoMLContainer.model_aliases.set(build_config.model_aliases)
         # This also verifies that svc can be imported correctly
         svc = import_service(
-            build_config.service, working_dir=build_ctx, standalone_load=True
+            build_config.service,
+            working_dir=build_ctx,
+            reload=reload,
+            standalone_load=True,
         )
         is_legacy = isinstance(svc, Service)
         # Apply default build options
@@ -229,13 +278,10 @@ def create(
         logger.debug(
             'Building BentoML service "%s" from build context "%s".', tag, build_ctx
         )
-
         bento_fs = TempFS(
             identifier=f"bentoml_bento_{bento_name}",
             temp_dir=BentoMLContainer.tmp_bento_store_dir.get(),
         )
-        ctx_fs = fs.open_fs(encode_path_for_uri(build_ctx))
-
         models: list[BentoModelInfo] = []
 
         def append_model(model: BentoModelInfo) -> None:
@@ -256,56 +302,59 @@ def append_model(model: BentoModelInfo) -> None:
                 for model in runner.models:
                     append_model(BentoModel(model.tag).to_info())
 
-        # create ignore specs
-        specs = BentoPathSpec(build_config.include, build_config.exclude, build_ctx)
+        if not bare:
+            ctx_fs = fs.open_fs(encode_path_for_uri(build_ctx))
 
-        # Copy all files base on include and exclude, into `src` directory
-        relpaths = [s for s in build_config.include if s.startswith("../")]
-        if len(relpaths) != 0:
-            raise InvalidArgument(
-                "Paths outside of the build context directory cannot be included; use a symlink or copy those files into the working directory manually."
-            )
-        bento_fs.makedir(BENTO_PROJECT_DIR_NAME)
-        target_fs = bento_fs.opendir(BENTO_PROJECT_DIR_NAME)
-        with target_fs.open(DEFAULT_BENTO_BUILD_FILE, "w") as bentofile_yaml:
-            build_config.to_yaml(bentofile_yaml)
-
-        for dir_path, _, files in ctx_fs.walk():
-            for f in files:
-                path = fs.path.combine(dir_path, f.name).lstrip("/")
-                if specs.includes(path):
-                    if ctx_fs.getsize(path) > 10 * 1024 * 1024:
-                        logger.warn("File size is larger than 10MiB: %s", path)
-                    target_fs.makedirs(dir_path, recreate=True)
-                    copy_file(ctx_fs, path, target_fs, path)
-
-        # NOTE: we need to generate both Python and Conda
-        # first to make sure we can generate the Dockerfile correctly.
-        build_config.python.write_to_bento(bento_fs, build_ctx, platform_=platform)
-        build_config.conda.write_to_bento(bento_fs, build_ctx)
-        build_config.docker.write_to_bento(bento_fs, build_ctx, build_config.conda)
-
-        # Create `readme.md` file
-        if build_config.description is None:
-            with bento_fs.open(BENTO_README_FILENAME, "w", encoding="utf-8") as f:
-                f.write(get_default_svc_readme(svc, svc_version=tag.version))
-        else:
-            if build_config.description.startswith("file:"):
-                file_name = build_config.description[5:].strip()
-                copy_file_to_fs_folder(
-                    file_name, bento_fs, dst_filename=BENTO_README_FILENAME
+            # create ignore specs
+            specs = BentoPathSpec(build_config.include, build_config.exclude, build_ctx)
+
+            # Copy all files base on include and exclude, into `src` directory
+            relpaths = [s for s in build_config.include if s.startswith("../")]
+            if len(relpaths) != 0:
+                raise InvalidArgument(
+                    "Paths outside of the build context directory cannot be included; use a symlink or copy those files into the working directory manually."
                 )
+            bento_fs.makedir(BENTO_PROJECT_DIR_NAME)
+            target_fs = bento_fs.opendir(BENTO_PROJECT_DIR_NAME)
+            with target_fs.open(DEFAULT_BENTO_BUILD_FILE, "w") as bentofile_yaml:
+                build_config.to_yaml(bentofile_yaml)
+
+            for dir_path, _, files in ctx_fs.walk():
+                for f in files:
+                    path = fs.path.combine(dir_path, f.name).lstrip("/")
+                    if specs.includes(path):
+                        if ctx_fs.getsize(path) > 10 * 1024 * 1024:
+                            logger.warn("File size is larger than 10MiB: %s", path)
+                        target_fs.makedirs(dir_path, recreate=True)
+                        copy_file(ctx_fs, path, target_fs, path)
+
+            # NOTE: we need to generate both Python and Conda
+            # first to make sure we can generate the Dockerfile correctly.
+            build_config.python.write_to_bento(bento_fs, build_ctx, platform_=platform)
+            build_config.conda.write_to_bento(bento_fs, build_ctx)
+            build_config.docker.write_to_bento(bento_fs, build_ctx, build_config.conda)
+
+            # Create `readme.md` file
+            if build_config.description is None:
+                with bento_fs.open(BENTO_README_FILENAME, "w", encoding="utf-8") as f:
+                    f.write(get_default_svc_readme(svc, svc_version=tag.version))
             else:
-                with bento_fs.open(BENTO_README_FILENAME, "w") as f:
-                    f.write(build_config.description)
-
-        # Create 'apis/openapi.yaml' file
-        bento_fs.makedir("apis")
-        with bento_fs.open(fs.path.combine("apis", "openapi.yaml"), "w") as f:
-            yaml.dump(svc.openapi_spec, f)
-        if not is_legacy:
-            with bento_fs.open(fs.path.combine("apis", "schema.json"), "w") as f:
-                json.dump(svc.schema(), f, indent=2)
+                if build_config.description.startswith("file:"):
+                    file_name = build_config.description[5:].strip()
+                    copy_file_to_fs_folder(
+                        file_name, bento_fs, dst_filename=BENTO_README_FILENAME
+                    )
+                else:
+                    with bento_fs.open(BENTO_README_FILENAME, "w") as f:
+                        f.write(build_config.description)
+
+            # Create 'apis/openapi.yaml' file
+            bento_fs.makedir("apis")
+            with bento_fs.open(fs.path.combine("apis", "openapi.yaml"), "w") as f:
+                yaml.dump(svc.openapi_spec, f)
+            if not is_legacy:
+                with bento_fs.open(fs.path.combine("apis", "schema.json"), "w") as f:
+                    json.dump(svc.schema(), f, indent=2)
 
         res = Bento(
             tag,
@@ -341,6 +390,8 @@ def append_model(model: BentoModelInfo) -> None:
                 schema=svc.schema() if not is_legacy else {},
             ),
         )
+        if bare:
+            return res
         # Create bento.yaml
         res.flush_info()
         try:
@@ -558,9 +609,9 @@ def from_bento_model(
 class BentoServiceInfo:
     name: str
     service: str
-    models: t.List[BentoModelInfo] = attr.field(factory=list)
-    dependencies: t.List[str] = attr.field(factory=list)
-    config: ServiceConfig = attr.field(factory=dict)
+    models: t.List[BentoModelInfo] = attr.field(factory=list, eq=False)
+    dependencies: t.List[str] = attr.field(factory=list, eq=False)
+    config: ServiceConfig = attr.field(factory=dict, eq=False)
 
     @classmethod
     def from_service(cls, svc: NewService[t.Any]) -> BentoServiceInfo:

src/bentoml/_internal/cloud/base.py

@@ -4,6 +4,7 @@
 from contextlib import contextmanager
 
 import attrs
+from rich import get_console
 from rich.console import Group
 from rich.live import Live
 from rich.panel import Panel
@@ -92,7 +93,8 @@ class Spinner:
     Use it as a context manager to start the live updating.
     """
 
-    def __init__(self):
+    def __init__(self, console: Console | None = None) -> None:
+        self.console = console or get_console()
         self.transmission_progress = Progress(
             TextColumn("[bold blue]{task.description}", justify="right"),
             BarColumn(bar_width=None),
@@ -103,6 +105,7 @@ def __init__(self):
             TransferSpeedColumn(),
             "•",
             TimeRemainingColumn(),
+            console=self.console,
         )
 
         self._logs: list[str] = []
@@ -111,24 +114,21 @@ def __init__(self):
             TimeElapsedColumn(),
             TextColumn("[bold purple]{task.description}"),
             SpinnerColumn("simpleDots"),
+            console=self.console,
         )
         self._spinner_task_id: t.Optional[TaskID] = None
-        self._live = Live(self)
-
-    @property
-    def console(self) -> "Console":
-        return self._live.console
+        self._live = Live(self, console=self.console)
 
     @contextmanager
     def spin(self, text: str) -> t.Generator[TaskID, None, None]:
         """Create a spinner as a context manager."""
+        task_id = self.update(text, new=True)
         try:
-            task_id = self.update(text, new=True)
             yield task_id
         finally:
-            self._spinner_task_id = None
-            self._spinner_progress.stop_task(task_id)
-            self._spinner_progress.update(task_id, visible=False)
+            self._spinner_progress.remove_task(task_id)
+            if self._spinner_task_id == task_id:
+                self._spinner_task_id = None
 
     def update(self, text: str, new: bool = False) -> TaskID:
         """Update the spin text."""
@@ -154,8 +154,7 @@ def start(self) -> None:
     def stop(self) -> None:
         """Stop live updating."""
         if self._spinner_task_id is not None:
-            self._spinner_progress.stop_task(self._spinner_task_id)
-            self._spinner_progress.update(self._spinner_task_id, visible=False)
+            self._spinner_progress.remove_task(self._spinner_task_id)
             self._spinner_task_id = None
         self._live.stop()
 

src/bentoml/_internal/cloud/bentocloud.py

@@ -65,13 +65,16 @@ def push_bento(
         bento: Bento,
         *,
         force: bool = False,
+        bare: bool = False,
         threads: int = 10,
     ):
         with self.spinner:
             upload_task_id = self.spinner.transmission_progress.add_task(
                 f'Pushing Bento "{bento.tag}"', start=False, visible=False
             )
-            self._do_push_bento(bento, upload_task_id, force=force, threads=threads)
+            self._do_push_bento(
+                bento, upload_task_id, force=force, threads=threads, bare=bare
+            )
             self.spinner.log(f'✅ Pushed Bento "{bento.tag}"')
 
     @inject
@@ -82,6 +85,7 @@ def _do_push_bento(
         *,
         force: bool = False,
         threads: int = 10,
+        bare: bool = False,
         rest_client: RestApiClient = Provide[BentoMLContainer.rest_api_client],
         bentoml_tmp_dir: str = Provide[BentoMLContainer.tmp_bento_store_dir],
     ):
@@ -115,6 +119,7 @@ def push_model(model: Model[t.Any]) -> None:
                 )
 
             executor.map(push_model, models_to_push)
+
         with self.spinner.spin(text=f'Fetching Bento repository "{name}"'):
             bento_repository = rest_client.v1.get_bento_repository(
                 bento_repository_name=name
@@ -200,7 +205,8 @@ def push_model(model: Model[t.Any]) -> None:
                         labels=labels,
                     ),
                 )
-
+        if bare:
+            return
         transmission_strategy: TransmissionStrategy = "proxy"
         presigned_upload_url: str | None = None