wandb 0.17.0__py3-none-win32.whl

wandb/testing/relay.py

@@ -0,0 +1,859 @@
+import dataclasses
+import json
+import logging
+import socket
+import sys
+import threading
+import traceback
+import urllib.parse
+from collections import defaultdict, deque
+from copy import deepcopy
+from typing import (
+    TYPE_CHECKING,
+    Any,
+    Callable,
+    Dict,
+    Iterable,
+    List,
+    Mapping,
+    Optional,
+    Union,
+)
+
+import flask
+import pandas as pd
+import requests
+import responses
+
+import wandb
+import wandb.util
+from wandb.sdk.lib.timer import Timer
+
+try:
+    from typing import Literal, TypedDict
+except ImportError:
+    from typing_extensions import Literal, TypedDict
+
+if sys.version_info >= (3, 8):
+    from typing import Protocol
+else:
+    from typing_extensions import Protocol
+
+if TYPE_CHECKING:
+    from typing import Deque
+
+    class RawRequestResponse(TypedDict):
+        url: str
+        request: Optional[Any]
+        response: Dict[str, Any]
+        time_elapsed: float  # seconds
+
+    ResolverName = Literal[
+        "upsert_bucket",
+        "upload_files",
+        "uploaded_files",
+        "preempting",
+        "upsert_sweep",
+    ]
+
+    class Resolver(TypedDict):
+        name: ResolverName
+        resolver: Callable[[Any], Optional[Dict[str, Any]]]
+
+
+class DeliberateHTTPError(Exception):
+    def __init__(self, message, status_code: int = 500):
+        Exception.__init__(self)
+        self.message = message
+        self.status_code = status_code
+
+    def get_response(self):
+        return flask.Response(self.message, status=self.status_code)
+
+    def __repr__(self):
+        return f"DeliberateHTTPError({self.message!r}, {self.status_code!r})"
+
+
+@dataclasses.dataclass
+class RunAttrs:
+    """Simple data class for run attributes."""
+
+    name: str
+    display_name: str
+    description: str
+    sweep_name: str
+    project: Dict[str, Any]
+    config: Dict[str, Any]
+    remote: Optional[str] = None
+    commit: Optional[str] = None
+
+
+class Context:
+    """A container used to store the snooped state/data of a test.
+
+    Includes raw requests and responses, parsed and processed data, and a number of
+    convenience methods and properties for accessing the data.
+    """
+
+    def __init__(self) -> None:
+        # parsed/merged data. keys are the individual wandb run id's.
+        self._entries = defaultdict(dict)
+        # container for raw requests and responses:
+        self.raw_data: List[RawRequestResponse] = []
+        # concatenated file contents for all runs:
+        self._history: Optional[pd.DataFrame] = None
+        self._events: Optional[pd.DataFrame] = None
+        self._summary: Optional[pd.DataFrame] = None
+        self._config: Optional[Dict[str, Any]] = None
+        self._output: Optional[Any] = None
+
+    def upsert(self, entry: Dict[str, Any]) -> None:
+        try:
+            entry_id: str = entry["name"]
+        except KeyError:
+            entry_id = entry["id"]
+        self._entries[entry_id] = wandb.util.merge_dicts(entry, self._entries[entry_id])
+
+    # mapping interface
+    def __getitem__(self, key: str) -> Any:
+        return self._entries[key]
+
+    def keys(self) -> Iterable[str]:
+        return self._entries.keys()
+
+    def get_file_contents(self, file_name: str) -> pd.DataFrame:
+        dfs = []
+
+        for entry_id in self._entries:
+            # - extract the content from `file_name`
+            # - sort by offset (will be useful when relay server goes async)
+            # - extract data, merge into a list of dicts and convert to a pandas dataframe
+            content_list = self._entries[entry_id].get("files", {}).get(file_name, [])
+            content_list.sort(key=lambda x: x["offset"])
+            content_list = [item["content"] for item in content_list]
+            # merge list of lists content_list:
+            content_list = [item for sublist in content_list for item in sublist]
+            df = pd.DataFrame.from_records(content_list)
+            df["__run_id"] = entry_id
+            dfs.append(df)
+
+        return pd.concat(dfs)
+
+    # attributes to use in assertions
+    @property
+    def entries(self) -> Dict[str, Any]:
+        return deepcopy(self._entries)
+
+    @property
+    def history(self) -> pd.DataFrame:
+        # todo: caveat: this assumes that all assertions happen at the end of a test
+        if self._history is not None:
+            return deepcopy(self._history)
+
+        self._history = self.get_file_contents("wandb-history.jsonl")
+        return deepcopy(self._history)
+
+    @property
+    def events(self) -> pd.DataFrame:
+        if self._events is not None:
+            return deepcopy(self._events)
+
+        self._events = self.get_file_contents("wandb-events.jsonl")
+        return deepcopy(self._events)
+
+    @property
+    def summary(self) -> pd.DataFrame:
+        if self._summary is not None:
+            return deepcopy(self._summary)
+
+        _summary = self.get_file_contents("wandb-summary.json")
+
+        # run summary may be updated multiple times,
+        # but we are only interested in the last one.
+        # we can have multiple runs saved to context,
+        # so we need to group by run id and take the
+        # last one for each run.
+        self._summary = (
+            _summary.groupby("__run_id").last().reset_index(level=["__run_id"])
+        )
+
+        return deepcopy(self._summary)
+
+    @property
+    def output(self) -> pd.DataFrame:
+        if self._output is not None:
+            return deepcopy(self._output)
+
+        self._output = self.get_file_contents("output.log")
+        return deepcopy(self._output)
+
+    @property
+    def config(self) -> Dict[str, Any]:
+        if self._config is not None:
+            return deepcopy(self._config)
+
+        self._config = {k: v["config"] for (k, v) in self._entries.items()}
+        return deepcopy(self._config)
+
+    # @property
+    # def telemetry(self) -> pd.DataFrame:
+    #     telemetry = pd.DataFrame.from_records(
+    #         [
+    #             {
+    #                 "__run_id": run_id,
+    #                 "telemetry": config.get("_wandb", {}).get("value", {}).get("t")
+    #             }
+    #             for (run_id, config) in self.config.items()
+    #         ]
+    #     )
+    #     return telemetry
+
+    # convenience data access methods
+    def get_run_telemetry(self, run_id: str) -> Dict[str, Any]:
+        return self.config.get(run_id, {}).get("_wandb", {}).get("value", {}).get("t")
+
+    def get_run_metrics(self, run_id: str) -> Dict[str, Any]:
+        return self.config.get(run_id, {}).get("_wandb", {}).get("value", {}).get("m")
+
+    def get_run_summary(
+        self, run_id: str, include_private: bool = False
+    ) -> Dict[str, Any]:
+        # run summary dataframe must have only one row
+        # for the given run id, so we convert it to dict
+        # and extract the first (and only) row.
+        mask_run = self.summary["__run_id"] == run_id
+        run_summary = self.summary[mask_run]
+        ret = (
+            run_summary.filter(regex="^[^_]", axis=1)
+            if not include_private
+            else run_summary
+        ).to_dict(orient="records")
+        return ret[0] if len(ret) > 0 else {}
+
+    def get_run_history(
+        self, run_id: str, include_private: bool = False
+    ) -> pd.DataFrame:
+        mask_run = self.history["__run_id"] == run_id
+        run_history = self.history[mask_run]
+        return (
+            run_history.filter(regex="^[^_]", axis=1)
+            if not include_private
+            else run_history
+        )
+
+    def get_run_uploaded_files(self, run_id: str) -> Dict[str, Any]:
+        return self.entries.get(run_id, {}).get("uploaded", [])
+
+    def get_run_stats(self, run_id: str) -> pd.DataFrame:
+        mask_run = self.events["__run_id"] == run_id
+        run_stats = self.events[mask_run]
+        return run_stats
+
+    def get_run_attrs(self, run_id: str) -> Optional[RunAttrs]:
+        run_entry = self._entries.get(run_id)
+        if not run_entry:
+            return None
+
+        return RunAttrs(
+            name=run_entry["name"],
+            display_name=run_entry["displayName"],
+            description=run_entry["description"],
+            sweep_name=run_entry["sweepName"],
+            project=run_entry["project"],
+            config=run_entry["config"],
+            remote=run_entry.get("repo"),
+            commit=run_entry.get("commit"),
+        )
+
+    def get_run(self, run_id: str) -> Dict[str, Any]:
+        return self._entries.get(run_id, {})
+
+    # todo: add getter (by run_id) utilities for other properties
+
+
+class QueryResolver:
+    """Resolve request/response pairs against a set of known patterns.
+
+    This extracts and processes useful data to be later stored in a Context object.
+    """
+
+    def __init__(self):
+        self.resolvers: List[Resolver] = [
+            {
+                "name": "upsert_bucket",
+                "resolver": self.resolve_upsert_bucket,
+            },
+            {
+                "name": "upload_files",
+                "resolver": self.resolve_upload_files,
+            },
+            {
+                "name": "uploaded_files",
+                "resolver": self.resolve_uploaded_files,
+            },
+            {
+                "name": "uploaded_files_legacy",
+                "resolver": self.resolve_uploaded_files_legacy,
+            },
+            {
+                "name": "preempting",
+                "resolver": self.resolve_preempting,
+            },
+            {
+                "name": "upsert_sweep",
+                "resolver": self.resolve_upsert_sweep,
+            },
+            {
+                "name": "create_artifact",
+                "resolver": self.resolve_create_artifact,
+            },
+            {
+                "name": "delete_run",
+                "resolver": self.resolve_delete_run,
+            },
+        ]
+
+    @staticmethod
+    def resolve_upsert_bucket(
+        request_data: Dict[str, Any], response_data: Dict[str, Any], **kwargs: Any
+    ) -> Optional[Dict[str, Any]]:
+        if not isinstance(request_data, dict) or not isinstance(response_data, dict):
+            return None
+        query = response_data.get("data", {}).get("upsertBucket") is not None
+        if query:
+            data = {
+                k: v for (k, v) in request_data["variables"].items() if v is not None
+            }
+            data.update(response_data["data"]["upsertBucket"].get("bucket"))
+            if "config" in data:
+                data["config"] = json.loads(data["config"])
+            return data
+        return None
+
+    @staticmethod
+    def resolve_delete_run(
+        request_data: Dict[str, Any], response_data: Dict[str, Any], **kwargs: Any
+    ) -> Optional[Dict[str, Any]]:
+        if not isinstance(request_data, dict) or not isinstance(response_data, dict):
+            return None
+        query = "query" in request_data and "deleteRun" in request_data["query"]
+        if query:
+            data = {
+                k: v for (k, v) in request_data["variables"].items() if v is not None
+            }
+            data.update(response_data["data"]["deleteRun"])
+            return data
+        return None
+
+    @staticmethod
+    def resolve_upload_files(
+        request_data: Dict[str, Any], response_data: Dict[str, Any], **kwargs: Any
+    ) -> Optional[Dict[str, Any]]:
+        if not isinstance(request_data, dict):
+            return None
+
+        query = request_data.get("files") is not None
+        if query:
+            # todo: refactor this 🤮🤮🤮🤮🤮 eventually?
+            name = kwargs.get("path").split("/")[2]
+            files = defaultdict(list)
+            for file_name, file_value in request_data["files"].items():
+                content = []
+                for k in file_value.get("content", []):
+                    try:
+                        content.append(json.loads(k))
+                    except json.decoder.JSONDecodeError:
+                        content.append([k])
+
+                files[file_name].append(
+                    {"offset": file_value.get("offset"), "content": content}
+                )
+
+            post_processed_data = {
+                "name": name,
+                "dropped": [request_data["dropped"]]
+                if "dropped" in request_data
+                else [],
+                "files": files,
+            }
+            return post_processed_data
+        return None
+
+    @staticmethod
+    def resolve_uploaded_files(
+        request_data: Dict[str, Any], response_data: Dict[str, Any], **kwargs: Any
+    ) -> Optional[Dict[str, Any]]:
+        if not isinstance(request_data, dict) or not isinstance(response_data, dict):
+            return None
+
+        query = "CreateRunFiles" in request_data.get("query", "")
+        if query:
+            run_name = request_data["variables"]["run"]
+            files = ((response_data.get("data") or {}).get("createRunFiles") or {}).get(
+                "files", {}
+            )
+            post_processed_data = {
+                "name": run_name,
+                "uploaded": [file["name"] for file in files] if files else [""],
+            }
+            return post_processed_data
+        return None
+
+    @staticmethod
+    def resolve_uploaded_files_legacy(
+        request_data: Dict[str, Any], response_data: Dict[str, Any], **kwargs: Any
+    ) -> Optional[Dict[str, Any]]:
+        # This is a legacy resolver for uploaded files
+        # No longer used by tests but leaving it here in case we need it in the future
+        # Please refer to upload_urls() in internal_api.py for more details
+        if not isinstance(request_data, dict) or not isinstance(response_data, dict):
+            return None
+
+        query = "RunUploadUrls" in request_data.get("query", "")
+        if query:
+            # todo: refactor this 🤮🤮🤮🤮🤮 eventually?
+            name = request_data["variables"]["run"]
+            files = (
+                response_data.get("data", {})
+                .get("model", {})
+                .get("bucket", {})
+                .get("files", {})
+                .get("edges", [])
+            )
+            # note: we count all attempts to upload files
+            post_processed_data = {
+                "name": name,
+                "uploaded": [files[0].get("node", {}).get("name")] if files else [""],
+            }
+            return post_processed_data
+        return None
+
+    @staticmethod
+    def resolve_preempting(
+        request_data: Dict[str, Any], response_data: Dict[str, Any], **kwargs: Any
+    ) -> Optional[Dict[str, Any]]:
+        if not isinstance(request_data, dict):
+            return None
+        query = "preempting" in request_data
+        if query:
+            name = kwargs.get("path").split("/")[2]
+            post_processed_data = {
+                "name": name,
+                "preempting": [request_data["preempting"]],
+            }
+            return post_processed_data
+        return None
+
+    @staticmethod
+    def resolve_upsert_sweep(
+        request_data: Dict[str, Any], response_data: Dict[str, Any], **kwargs: Any
+    ) -> Optional[Dict[str, Any]]:
+        if not isinstance(response_data, dict):
+            return None
+        query = response_data.get("data", {}).get("upsertSweep") is not None
+        if query:
+            data = response_data["data"]["upsertSweep"].get("sweep")
+            return data
+        return None
+
+    def resolve_create_artifact(
+        self, request_data: Dict[str, Any], response_data: Dict[str, Any], **kwargs: Any
+    ) -> Optional[Dict[str, Any]]:
+        if not isinstance(request_data, dict):
+            return None
+        query = (
+            "createArtifact(" in request_data.get("query", "")
+            and request_data.get("variables") is not None
+            and response_data is not None
+        )
+        if query:
+            name = request_data["variables"]["runName"]
+            post_processed_data = {
+                "name": name,
+                "create_artifact": [
+                    {
+                        "variables": request_data["variables"],
+                        "response": response_data["data"]["createArtifact"]["artifact"],
+                    }
+                ],
+            }
+            return post_processed_data
+        return None
+
+    def resolve(
+        self,
+        request_data: Dict[str, Any],
+        response_data: Dict[str, Any],
+        **kwargs: Any,
+    ) -> Optional[Dict[str, Any]]:
+        results = []
+        for resolver in self.resolvers:
+            result = resolver.get("resolver")(request_data, response_data, **kwargs)
+            if result is not None:
+                results.append(result)
+        return results
+
+
+class TokenizedCircularPattern:
+    APPLY_TOKEN = "1"
+    PASS_TOKEN = "0"
+    STOP_TOKEN = "2"
+
+    def __init__(self, pattern: str):
+        known_tokens = {self.APPLY_TOKEN, self.PASS_TOKEN, self.STOP_TOKEN}
+        if not pattern:
+            raise ValueError("Pattern cannot be empty")
+
+        if set(pattern) - known_tokens:
+            raise ValueError(f"Pattern can only contain {known_tokens}")
+        self.pattern: Deque[str] = deque(pattern)
+
+    def next(self):
+        if self.pattern[0] == self.STOP_TOKEN:
+            return
+        self.pattern.rotate(-1)
+
+    def should_apply(self) -> bool:
+        return self.pattern[0] == self.APPLY_TOKEN
+
+
+@dataclasses.dataclass
+class InjectedResponse:
+    method: str
+    url: str
+    body: Union[str, Exception]
+    status: int = 200
+    content_type: str = "text/plain"
+    # todo: add more fields for other types of responses?
+    custom_match_fn: Optional[Callable[..., bool]] = None
+    application_pattern: TokenizedCircularPattern = TokenizedCircularPattern("1")
+
+    # application_pattern defines the pattern of the response injection
+    # as the requests come in.
+    # 0 == do not inject the response
+    # 1 == inject the response
+    # 2 == stop using the response (END token)
+    #
+    # - when no END token is present, the pattern is repeated indefinitely
+    # - when END token is present, the pattern is applied until the END token is reached
+    # - to replicate the current behavior:
+    #  - use application_pattern = "1" if wanting to apply the pattern to all requests
+    #  - use application_pattern = "1" * COUNTER + "2" to apply the pattern to the first COUNTER requests
+    #
+    # Examples of application_pattern:
+    # 1. application_pattern = "1012"
+    #    - inject the response for the first request
+    #    - do not inject the response for the second request
+    #    - inject the response for the third request
+    #    - stop using the response starting from the fourth request onwards
+    # 2. application_pattern = "110"
+    #    repeat the following pattern indefinitely:
+    #    - inject the response for the first request
+    #    - inject the response for the second request
+    #    - stop using the response for the third request
+
+    def __eq__(
+        self,
+        other: Union["InjectedResponse", requests.Request, requests.PreparedRequest],
+    ):
+        """Check InjectedResponse object equality.
+
+        We use this to check if this response should be injected as a replacement of
+        `other`.
+
+        :param other:
+        :return:
+        """
+        if not isinstance(
+            other, (InjectedResponse, requests.Request, requests.PreparedRequest)
+        ):
+            return False
+
+        # always check the method and url
+        ret = self.method == other.method and self.url == other.url
+        # use custom_match_fn to check, e.g. the request body content
+        if ret and self.custom_match_fn is not None:
+            ret = self.custom_match_fn(self, other)
+        return ret
+
+    def to_dict(self):
+        excluded_fields = {"application_pattern", "custom_match_fn"}
+        return {
+            k: self.__getattribute__(k)
+            for k in self.__dict__
+            if (not k.startswith("_") and k not in excluded_fields)
+        }
+
+
+class RelayControlProtocol(Protocol):
+    def process(self, request: "flask.Request") -> None: ...  # pragma: no cover
+
+    def control(
+        self, request: "flask.Request"
+    ) -> Mapping[str, str]: ...  # pragma: no cover
+
+
+class RelayServer:
+    def __init__(
+        self,
+        base_url: str,
+        inject: Optional[List[InjectedResponse]] = None,
+        control: Optional[RelayControlProtocol] = None,
+        verbose: bool = False,
+    ) -> None:
+        # todo for the future:
+        #  - consider switching from Flask to Quart
+        #  - async app will allow for better failure injection/poor network perf
+        self.relay_control = control
+        self.app = flask.Flask(__name__)
+        self.app.logger.setLevel(logging.INFO)
+        self.app.register_error_handler(DeliberateHTTPError, self.handle_http_exception)
+        self.app.add_url_rule(
+            rule="/graphql",
+            endpoint="graphql",
+            view_func=self.graphql,
+            methods=["POST"],
+        )
+        self.app.add_url_rule(
+            rule="/files/<path:path>",
+            endpoint="files",
+            view_func=self.file_stream,
+            methods=["POST"],
+        )
+        self.app.add_url_rule(
+            rule="/storage",
+            endpoint="storage",
+            view_func=self.storage,
+            methods=["PUT", "GET"],
+        )
+        self.app.add_url_rule(
+            rule="/storage/<path:path>",
+            endpoint="storage_file",
+            view_func=self.storage_file,
+            methods=["PUT", "GET"],
+        )
+        if control:
+            self.app.add_url_rule(
+                rule="/_control",
+                endpoint="_control",
+                view_func=self.control,
+                methods=["POST"],
+            )
+        # @app.route("/artifacts/<entity>/<digest>", methods=["GET", "POST"])
+        self.port = self._get_free_port()
+        self.base_url = urllib.parse.urlparse(base_url)
+        self.session = requests.Session()
+        self.relay_url = f"http://127.0.0.1:{self.port}"
+
+        # todo: add an option to add custom resolvers
+        self.resolver = QueryResolver()
+        # recursively merge-able object to store state
+        self.context = Context()
+
+        # injected responses
+        self.inject = inject or []
+
+        # useful when debugging:
+        # self.after_request_fn = self.app.after_request(self.after_request_fn)
+        self.verbose = verbose
+
+    @staticmethod
+    def handle_http_exception(e):
+        response = e.get_response()
+        return response
+
+    @staticmethod
+    def _get_free_port() -> int:
+        sock = socket.socket()
+        sock.bind(("", 0))
+
+        _, port = sock.getsockname()
+        return port
+
+    def start(self) -> None:
+        # run server in a separate thread
+        relay_server_thread = threading.Thread(
+            target=self.app.run,
+            kwargs={"port": self.port},
+            daemon=True,
+        )
+        relay_server_thread.start()
+
+    def after_request_fn(self, response: "requests.Response") -> "requests.Response":
+        # todo: this is useful for debugging, but should be removed in the future
+        # flask.request.url = self.relay_url + flask.request.url
+        print(flask.request)
+        print(flask.request.get_json())
+        print(response)
+        print(response.json())
+        return response
+
+    def relay(
+        self,
+        request: "flask.Request",
+    ) -> Union["responses.Response", "requests.Response"]:
+        # replace the relay url with the real backend url (self.base_url)
+        url = (
+            urllib.parse.urlparse(request.url)
+            ._replace(netloc=self.base_url.netloc, scheme=self.base_url.scheme)
+            .geturl()
+        )
+        headers = {key: value for (key, value) in request.headers if key != "Host"}
+        prepared_relayed_request = requests.Request(
+            method=request.method,
+            url=url,
+            headers=headers,
+            data=request.get_data(),
+            json=request.get_json(),
+        ).prepare()
+
+        if self.verbose:
+            print("*****************")
+            print("RELAY REQUEST:")
+            print(prepared_relayed_request.url)
+            print(prepared_relayed_request.method)
+            print(prepared_relayed_request.headers)
+            print(prepared_relayed_request.body)
+            print("*****************")
+
+        for injected_response in self.inject:
+            # where are we in the application pattern?
+            should_apply = injected_response.application_pattern.should_apply()
+            # check if an injected response matches the request
+            if injected_response == prepared_relayed_request:
+                if self.verbose:
+                    print("*****************")
+                    print("INJECTING RESPONSE:")
+                    print(injected_response.to_dict())
+                    print("*****************")
+                # rotate the injection pattern
+                injected_response.application_pattern.next()
+                # TODO: allow access to the request object when making the mocked response
+                if should_apply:
+                    with responses.RequestsMock() as mocked_responses:
+                        # do the actual injection
+                        mocked_responses.add(**injected_response.to_dict())
+                        relayed_response = self.session.send(prepared_relayed_request)
+
+                        return relayed_response
+
+        # normal case: no injected response matches the request
+        relayed_response = self.session.send(prepared_relayed_request)
+        return relayed_response
+
+    def snoop_context(
+        self,
+        request: "flask.Request",
+        response: "requests.Response",
+        time_elapsed: float,
+        **kwargs: Any,
+    ) -> None:
+        request_data = request.get_json()
+        response_data = response.json() or {}
+
+        if self.relay_control:
+            self.relay_control.process(request)
+
+        # store raw data
+        raw_data: RawRequestResponse = {
+            "url": request.url,
+            "request": request_data,
+            "response": response_data,
+            "time_elapsed": time_elapsed,
+        }
+        self.context.raw_data.append(raw_data)
+
+        try:
+            snooped_context = self.resolver.resolve(
+                request_data,
+                response_data,
+                **kwargs,
+            )
+            for entry in snooped_context:
+                self.context.upsert(entry)
+        except Exception as e:
+            print("Failed to resolve context: ", e)
+            traceback.print_exc()
+            snooped_context = None
+
+        return None
+
+    def graphql(self) -> Mapping[str, str]:
+        request = flask.request
+        with Timer() as timer:
+            relayed_response = self.relay(request)
+        if self.verbose:
+            print("*****************")
+            print("GRAPHQL REQUEST:")
+            print(request.get_json())
+            print("GRAPHQL RESPONSE:")
+            print(relayed_response.status_code, relayed_response.json())
+            print("*****************")
+        # snoop work to extract the context
+        self.snoop_context(request, relayed_response, timer.elapsed)
+        if self.verbose:
+            print("*****************")
+            print("SNOOPED CONTEXT:")
+            print(self.context.entries)
+            print(len(self.context.raw_data))
+            print("*****************")
+
+        return relayed_response.json()
+
+    def file_stream(self, path) -> Mapping[str, str]:
+        request = flask.request
+        with Timer() as timer:
+            relayed_response = self.relay(request)
+        if self.verbose:
+            print("*****************")
+            print("FILE STREAM REQUEST:")
+            print("********PATH*********")
+            print(path)
+            print("********ENDPATH*********")
+            print(request.get_json())
+            print("FILE STREAM RESPONSE:")
+            print(relayed_response)
+            print(relayed_response.status_code, relayed_response.json())
+            print("*****************")
+        self.snoop_context(request, relayed_response, timer.elapsed, path=path)
+
+        return relayed_response.json()
+
+    def storage(self) -> Mapping[str, str]:
+        request = flask.request
+        with Timer() as timer:
+            relayed_response = self.relay(request)
+        if self.verbose:
+            print("*****************")
+            print("STORAGE REQUEST:")
+            print(request.get_json())
+            print("STORAGE RESPONSE:")
+            print(relayed_response.status_code, relayed_response.json())
+            print("*****************")
+
+        self.snoop_context(request, relayed_response, timer.elapsed)
+
+        return relayed_response.json()
+
+    def storage_file(self, path) -> Mapping[str, str]:
+        request = flask.request
+        with Timer() as timer:
+            relayed_response = self.relay(request)
+        if self.verbose:
+            print("*****************")
+            print("STORAGE FILE REQUEST:")
+            print("********PATH*********")
+            print(path)
+            print("********ENDPATH*********")
+            print(request.get_json())
+            print("STORAGE FILE RESPONSE:")
+            print(relayed_response.json())
+            print("*****************")
+
+        self.snoop_context(request, relayed_response, timer.elapsed, path=path)
+
+        return relayed_response.json()
+
+    def control(self) -> Mapping[str, str]:
+        assert self.relay_control
+        return self.relay_control.control(flask.request)