feat(bentocloud): deployment v2 api client + cli

[Haivilo]

What does this PR address?

bentoml.deployment.create(...)
bentoml.deployment.update(...)
bentoml.deployment.apply(...)
bentoml.deployment.get(...)
bentoml.deployment.list(...)
bentoml.deployment.terminate(...)
bentoml.deployment.delete(...)

bentoml deploy .
bentoml deployment update
bentoml deployment get
bentoml deployment list
bentoml deployment terminate
bentoml deployment delete

Before submitting:

• Does the Pull Request follow Conventional Commits specification naming? Here are GitHub's
  guide on how to create a pull request.
• Does the code follow BentoML's code style, pre-commit run -a script has passed (instructions)?
• Did you read through contribution guidelines and follow development guidelines?
• Did your changes require updates to the documentation? Have you updated
  those accordingly? Here are documentation guidelines and tips on writting docs.
• Did you write tests to cover your changes?

[frostming]

What about making get_rest_api_client accept a version parameter instead of accessing via client.v1?

def get_rest_api_client(version = "v2"):
    ...
    return getattr(client, version)

[Haivilo]

What about making get_rest_api_client accept a version parameter instead of accessing via client.v1?

def get_rest_api_client(version = "v2"):
    ...
    return getattr(client, version)

I am not sure if we should do this, because some methods use both v1 and v2. If we were to do it this way, then we need to instantiate it in 2 lines instead of 1. Is there any draw back to return the whole client?

client_v1 = get_rest_api_client(version = "v1")
client_v2 = get_rest_api_client(version = "v2")
...

[aarnphm]

What about making get_rest_api_client accept a version parameter instead of accessing via client.v1?

def get_rest_api_client(version = "v2"):
    ...
    return getattr(client, version)

I am not sure if we should do this, because some methods use both v1 and v2. If we were to do it this way, then we need to instantiate it in 2 lines instead of 1. Is there any draw back to return the whole client?

client_v1 = get_rest_api_client(version = "v1")
client_v2 = get_rest_api_client(version = "v2")
...

This get_rest_api_client is internal right?

Why do we need v1 still?

[Haivilo]

What about making get_rest_api_client accept a version parameter instead of accessing via client.v1?

def get_rest_api_client(version = "v2"):
    ...
    return getattr(client, version)

I am not sure if we should do this, because some methods use both v1 and v2. If we were to do it this way, then we need to instantiate it in 2 lines instead of 1. Is there any draw back to return the whole client?

client_v1 = get_rest_api_client(version = "v1")
client_v2 = get_rest_api_client(version = "v2")
...

This get_rest_api_client is internal right?

Why do we need v1 still?

Only update and create deployment has v2 api. We will be using v1 api for all the rest of the endpoints:

BentoML/src/bentoml/_internal/cloud/client.py

Lines 539 to 555 in af2a794

	class RestApiClientV2(BaseRestApiClient):
	def create_deployment(
	self, create_schema: CreateDeploymentSchemaV2
	) -> DeploymentFullSchema | None:
	url = urljoin(self.endpoint, "/api/v2/deployments")
	resp = self.session.post(url, data=schema_to_json(create_schema))
	self._check_resp(resp)
	return schema_from_json(resp.text, DeploymentFullSchema)
	def update_deployment(
	self, deployment_name: str, update_schema: UpdateDeploymentSchemaV2
	) -> DeploymentFullSchema | None:
	url = urljoin(self.endpoint, f"/api/v2/deployments/{deployment_name}")
	resp = self.session.post(url, data=schema_to_json(update_schema))
	self._check_resp(resp)
	return schema_from_json(resp.text, DeploymentFullSchema)

[aarnphm]

What about making get_rest_api_client accept a version parameter instead of accessing via client.v1?

def get_rest_api_client(version = "v2"):
    ...
    return getattr(client, version)

I am not sure if we should do this, because some methods use both v1 and v2. If we were to do it this way, then we need to instantiate it in 2 lines instead of 1. Is there any draw back to return the whole client?

client_v1 = get_rest_api_client(version = "v1")
client_v2 = get_rest_api_client(version = "v2")
...

This get_rest_api_client is internal right?
Why do we need v1 still?

Only update and create deployment has v2 api. We will be using v1 api for all the rest of the endpoints:

BentoML/src/bentoml/_internal/cloud/client.py

Lines 539 to 555 in af2a794

	class RestApiClientV2(BaseRestApiClient):
	def create_deployment(
	self, create_schema: CreateDeploymentSchemaV2
	) -> DeploymentFullSchema | None:
	url = urljoin(self.endpoint, "/api/v2/deployments")
	resp = self.session.post(url, data=schema_to_json(create_schema))
	self._check_resp(resp)
	return schema_from_json(resp.text, DeploymentFullSchema)
	def update_deployment(
	self, deployment_name: str, update_schema: UpdateDeploymentSchemaV2
	) -> DeploymentFullSchema | None:
	url = urljoin(self.endpoint, f"/api/v2/deployments/{deployment_name}")
	resp = self.session.post(url, data=schema_to_json(update_schema))
	self._check_resp(resp)
	return schema_from_json(resp.text, DeploymentFullSchema)

Hmm I propose this: for v2 client it would just extends v1. So only update and create deployments is different. Otherwise it will still be the same.

[Haivilo]

Discussed with @aarnphm and @frostming , let's keep using client.v1.xxx and client.v2.xxx since it's internal.

[aarnphm]

A few final comments. Then we can get this in

[aarnphm]

Few things.

[aarnphm]

Note to @aarnphm: Here once migrate to tqdm.

[aarnphm]

LGTM

[aarnphm]

LGTM