feat(agent/workspace): Add GCS and S3 FileWorkspace providers (#6485)

* refactor: Rename FileWorkspace to LocalFileWorkspace and create FileWorkspace abstract class
  - Rename `FileWorkspace` to `LocalFileWorkspace` to provide a more descriptive name for the class that represents a file workspace that works with local files.
  - Create a new base class `FileWorkspace` to serve as the parent class for `LocalFileWorkspace`. This allows for easier extension and customization of file workspaces in the future.
  - Update import statements and references to `FileWorkspace` throughout the codebase to use the new naming conventions.

* feat: Add S3FileWorkspace + tests + test setups for CI and Docker
  - Added S3FileWorkspace class to provide an interface for interacting with a file workspace and storing files in an S3 bucket.
  - Updated pyproject.toml to include dependencies for boto3 and boto3-stubs.
  - Implemented unit tests for S3FileWorkspace.
  - Added MinIO service to Docker CI to allow testing S3 features in CI.
  - Added autogpt-test service config to docker-compose.yml for local testing with MinIO.

* ci(docker): tee test output instead of capturing

* fix: Improve error handling in S3FileWorkspace.initialize()
  - Do not tolerate all `botocore.exceptions.ClientError`s
  - Raise the exception anyways if the error is not "NoSuchBucket"

* feat: Add S3 workspace backend support and S3Credentials
  - Added support for S3 workspace backend in the Autogpt configuration
  - Added a new sub-config `S3Credentials` to store S3 credentials
  - Modified the `.env.template` file to include variables related to S3 credentials
  - Added a new `s3_credentials` attribute on the `Config` class to store S3 credentials
  - Moved the `unmasked` method from `ModelProviderCredentials` to the parent `ProviderCredentials` class to handle unmasking for S3 credentials

* fix(agent/tests): Fix S3FileWorkspace initialization in test_s3_file_workspace.py
  - Update the S3FileWorkspace initialization in the test_s3_file_workspace.py file to include the required S3 Credentials.

* refactor: Remove S3Credentials and add get_workspace function
  - Remove `S3Credentials` as boto3 will fetch the config from the environment by itself
  - Add `get_workspace` function in `autogpt.file_workspace` module
  - Update `.env.template` and tests to reflect the changes

* feat(agent/workspace): Make agent workspace backend configurable
  - Modified `autogpt.file_workspace.get_workspace` function to either take a workspace `id` or `root_path`.
  - Modified `FileWorkspaceMixin` to use the `get_workspace` function to set up the workspace.
  - Updated the type hints and imports accordingly.

* feat(agent/workspace): Add GCSFileWorkspace for Google Cloud Storage
  - Added support for Google Cloud Storage as a storage backend option in the workspace.
  - Created the `GCSFileWorkspace` class to interface with a file workspace stored in a Google Cloud Storage bucket.
  - Implemented the `GCSFileWorkspaceConfiguration` class to handle the configuration for Google Cloud Storage workspaces.
  - Updated the `get_workspace` function to include the option to use Google Cloud Storage as a workspace backend.
  - Added unit tests for the new `GCSFileWorkspace` class.

* fix: Unbreak use of non-local workspaces in AgentProtocolServer
  - Modify the `_get_task_agent_file_workspace` method to handle both local and non-local workspaces correctly

.github/workflows/autogpt-ci.yml

@@ -83,6 +83,15 @@ jobs:
       matrix:
         python-version: ["3.10"]
 
+    services:
+      minio:
+        image: minio/minio:edge-cicd
+        ports:
+          - 9000:9000
+        options: >
+          --health-interval=10s --health-timeout=5s --health-retries=3
+          --health-cmd="curl -f http://localhost:9000/minio/health/live"
+
     steps:
       - name: Checkout repository
         uses: actions/checkout@v3
@@ -154,8 +163,11 @@ jobs:
             tests/unit tests/integration
         env:
           CI: true
-          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
           PLAIN_OUTPUT: True
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          S3_ENDPOINT_URL: http://localhost:9000
+          AWS_ACCESS_KEY_ID: minioadmin
+          AWS_SECRET_ACCESS_KEY: minioadmin
 
       - name: Upload coverage reports to Codecov
         uses: codecov/codecov-action@v3

.github/workflows/autogpt-docker-ci.yml

@@ -89,6 +89,15 @@ jobs:
   test:
     runs-on: ubuntu-latest
     timeout-minutes: 10
+
+    services:
+      minio:
+        image: minio/minio:edge-cicd
+        options: >
+          --name=minio
+          --health-interval=10s --health-timeout=5s --health-retries=3
+          --health-cmd="curl -f http://localhost:9000/minio/health/live"
+
     steps:
       - name: Check out repository
         uses: actions/checkout@v3
@@ -124,23 +133,25 @@ jobs:
           CI: true
           PLAIN_OUTPUT: True
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          S3_ENDPOINT_URL: http://minio:9000
+          AWS_ACCESS_KEY_ID: minioadmin
+          AWS_SECRET_ACCESS_KEY: minioadmin
         run: |
           set +e
-          test_output=$(
-            docker run --env CI --env OPENAI_API_KEY \
-              --entrypoint poetry ${{ env.IMAGE_NAME }} run \
-              pytest -v --cov=autogpt --cov-branch --cov-report term-missing \
-              --numprocesses=4 --durations=10 \
-              tests/unit tests/integration 2>&1
-          )
-          test_failure=$?
+          docker run --env CI --env OPENAI_API_KEY \
+            --network container:minio \
+            --env S3_ENDPOINT_URL --env AWS_ACCESS_KEY_ID --env AWS_SECRET_ACCESS_KEY \
+            --entrypoint poetry ${{ env.IMAGE_NAME }} run \
+            pytest -v --cov=autogpt --cov-branch --cov-report term-missing \
+            --numprocesses=4 --durations=10 \
+            tests/unit tests/integration 2>&1 | tee test_output.txt
 
-          echo "$test_output"
+          test_failure=${PIPESTATUS[0]}
 
           cat << $EOF >> $GITHUB_STEP_SUMMARY
           # Tests $([ $test_failure = 0 ] && echo '✅' || echo '❌')
           \`\`\`
-          $test_output
+          $(cat test_output.txt)
           \`\`\`
           $EOF
 

autogpts/autogpt/.env.template

@@ -8,9 +8,32 @@ OPENAI_API_KEY=your-openai-api-key
 ## EXECUTE_LOCAL_COMMANDS - Allow local command execution (Default: False)
 # EXECUTE_LOCAL_COMMANDS=False
 
+### Workspace ###
+
 ## RESTRICT_TO_WORKSPACE - Restrict file operations to workspace ./data/agents/<agent_id>/workspace (Default: True)
 # RESTRICT_TO_WORKSPACE=True
 
+## DISABLED_COMMAND_CATEGORIES - The list of categories of commands that are disabled (Default: None)
+# DISABLED_COMMAND_CATEGORIES=
+
+## WORKSPACE_BACKEND - Choose a storage backend for workspace contents
+## Options: local, gcs, s3
+# WORKSPACE_BACKEND=local
+
+## WORKSPACE_STORAGE_BUCKET - GCS/S3 Bucket to store workspace contents in
+# WORKSPACE_STORAGE_BUCKET=autogpt
+
+## GCS Credentials
+# see https://cloud.google.com/storage/docs/authentication#libauth
+
+## AWS/S3 Credentials
+# see https://boto3.amazonaws.com/v1/documentation/api/latest/guide/configuration.html
+
+## S3_ENDPOINT_URL - If you're using non-AWS S3, set your endpoint here.
+# S3_ENDPOINT_URL=
+
+### Miscellaneous ###
+
 ## USER_AGENT - Define the user-agent used by the requests library to browse website (string)
 # USER_AGENT="Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_4) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/83.0.4103.97 Safari/537.36"
 
@@ -29,12 +52,6 @@ OPENAI_API_KEY=your-openai-api-key
 ## EXIT_KEY - Key to exit AutoGPT
 # EXIT_KEY=n
 
-## PLAIN_OUTPUT - Plain output, which disables the spinner (Default: False)
-# PLAIN_OUTPUT=False
-
-## DISABLED_COMMAND_CATEGORIES - The list of categories of commands that are disabled (Default: None)
-# DISABLED_COMMAND_CATEGORIES=
-
 ################################################################################
 ### LLM PROVIDER
 ################################################################################
@@ -201,5 +218,5 @@ OPENAI_API_KEY=your-openai-api-key
 ## Note: Log file output is disabled if LOG_FORMAT=structured_google_cloud.
 # LOG_FILE_FORMAT=simple
 
-## PLAIN_OUTPUT - Disables animated typing in the console output.
+## PLAIN_OUTPUT - Disables animated typing and the spinner in the console output. (Default: False)
 # PLAIN_OUTPUT=False

autogpts/autogpt/autogpt/agents/features/file_workspace.py

@@ -5,11 +5,15 @@ from typing import TYPE_CHECKING
 if TYPE_CHECKING:
     from pathlib import Path
 
-    from ..base import BaseAgent
+    from ..base import BaseAgent, Config
 
-from autogpt.file_workspace import FileWorkspace
+from autogpt.file_workspace import (
+    FileWorkspace,
+    FileWorkspaceBackendName,
+    get_workspace,
+)
 
-from ..base import AgentFileManager, BaseAgentConfiguration
+from ..base import AgentFileManager, BaseAgentSettings
 
 
 class FileWorkspaceMixin:
@@ -22,32 +26,36 @@ class FileWorkspaceMixin:
         # Initialize other bases first, because we need the config from BaseAgent
         super(FileWorkspaceMixin, self).__init__(**kwargs)
 
-        config: BaseAgentConfiguration = getattr(self, "config")
-        if not isinstance(config, BaseAgentConfiguration):
-            raise ValueError(
-                "Cannot initialize Workspace for Agent without compatible .config"
-            )
         file_manager: AgentFileManager = getattr(self, "file_manager")
         if not file_manager:
             return
 
-        self.workspace = _setup_workspace(file_manager, config)
+        self._setup_workspace()
 
     def attach_fs(self, agent_dir: Path):
         res = super(FileWorkspaceMixin, self).attach_fs(agent_dir)
 
-        self.workspace = _setup_workspace(self.file_manager, self.config)
+        self._setup_workspace()
 
         return res
 
+    def _setup_workspace(self) -> None:
+        settings: BaseAgentSettings = getattr(self, "state")
+        assert settings.agent_id, "Cannot attach workspace to anonymous agent"
+        app_config: Config = getattr(self, "legacy_config")
+        file_manager: AgentFileManager = getattr(self, "file_manager")
 
-def _setup_workspace(file_manager: AgentFileManager, config: BaseAgentConfiguration):
-    workspace = FileWorkspace(
-        file_manager.root / "workspace",
-        restrict_to_root=not config.allow_fs_access,
-    )
-    workspace.initialize()
-    return workspace
+        ws_backend = app_config.workspace_backend
+        local = ws_backend == FileWorkspaceBackendName.LOCAL
+        workspace = get_workspace(
+            backend=ws_backend,
+            id=settings.agent_id if not local else "",
+            root_path=file_manager.root / "workspace" if local else None,
+        )
+        if local and settings.config.allow_fs_access:
+            workspace._restrict_to_root = False  # type: ignore
+        workspace.initialize()
+        self.workspace = workspace
 
 
 def get_agent_workspace(agent: BaseAgent) -> FileWorkspace | None:

autogpts/autogpt/autogpt/app/agent_protocol_server.py

@@ -33,7 +33,11 @@ from autogpt.commands.system import finish
 from autogpt.commands.user_interaction import ask_user
 from autogpt.config import Config
 from autogpt.core.resource.model_providers import ChatModelProvider
-from autogpt.file_workspace import FileWorkspace
+from autogpt.file_workspace import (
+    FileWorkspace,
+    FileWorkspaceBackendName,
+    get_workspace,
+)
 from autogpt.models.action_history import ActionErrorResult, ActionSuccessResult
 
 logger = logging.getLogger(__name__)
@@ -340,7 +344,7 @@ class AgentProtocolServer:
         else:
             file_path = os.path.join(relative_path, file_name)
 
-        workspace = get_task_agent_file_workspace(task_id, self.agent_manager)
+        workspace = self._get_task_agent_file_workspace(task_id, self.agent_manager)
         await workspace.write_file(file_path, data)
 
         artifact = await self.db.create_artifact(
@@ -361,7 +365,7 @@ class AgentProtocolServer:
                 file_path = os.path.join(artifact.relative_path, artifact.file_name)
             else:
                 file_path = artifact.relative_path
-            workspace = get_task_agent_file_workspace(task_id, self.agent_manager)
+            workspace = self._get_task_agent_file_workspace(task_id, self.agent_manager)
             retrieved_artifact = workspace.read_file(file_path, binary=True)
         except NotFoundError:
             raise
@@ -376,24 +380,33 @@ class AgentProtocolServer:
             },
         )
 
+    def _get_task_agent_file_workspace(
+        self,
+        task_id: str | int,
+        agent_manager: AgentManager,
+    ) -> FileWorkspace:
+        use_local_ws = (
+            self.app_config.workspace_backend == FileWorkspaceBackendName.LOCAL
+        )
+        agent_id = task_agent_id(task_id)
+        workspace = get_workspace(
+            backend=self.app_config.workspace_backend,
+            id=agent_id if not use_local_ws else "",
+            root_path=agent_manager.get_agent_dir(
+                agent_id=agent_id,
+                must_exist=True,
+            )
+            / "workspace"
+            if use_local_ws
+            else None,
+        )
+        workspace.initialize()
+        return workspace
+
 
 def task_agent_id(task_id: str | int) -> str:
     return f"AutoGPT-{task_id}"
 
 
-def get_task_agent_file_workspace(
-    task_id: str | int,
-    agent_manager: AgentManager,
-) -> FileWorkspace:
-    return FileWorkspace(
-        root=agent_manager.get_agent_dir(
-            agent_id=task_agent_id(task_id),
-            must_exist=True,
-        )
-        / "workspace",
-        restrict_to_root=True,
-    )
-
-
 def fmt_kwargs(kwargs: dict) -> str:
     return ", ".join(f"{n}={repr(v)}" for n, v in kwargs.items())

autogpts/autogpt/autogpt/config/config.py

@@ -20,6 +20,7 @@ from autogpt.core.resource.model_providers.openai import (
     OPEN_AI_CHAT_MODELS,
     OpenAICredentials,
 )
+from autogpt.file_workspace import FileWorkspaceBackendName
 from autogpt.logs.config import LoggingConfig
 from autogpt.plugins.plugins_config import PluginsConfig
 from autogpt.speech import TTSConfig
@@ -51,10 +52,19 @@ class Config(SystemSettings, arbitrary_types_allowed=True):
     chat_messages_enabled: bool = UserConfigurable(
         default=True, from_env=lambda: os.getenv("CHAT_MESSAGES_ENABLED") == "True"
     )
+
     # TTS configuration
     tts_config: TTSConfig = TTSConfig()
     logging: LoggingConfig = LoggingConfig()
 
+    # Workspace
+    workspace_backend: FileWorkspaceBackendName = UserConfigurable(
+        default=FileWorkspaceBackendName.LOCAL,
+        from_env=lambda: FileWorkspaceBackendName(v)
+        if (v := os.getenv("WORKSPACE_BACKEND"))
+        else None,
+    )
+
     ##########################
     # Agent Control Settings #
     ##########################

autogpts/autogpt/autogpt/core/resource/model_providers/schema.py

@@ -172,24 +172,10 @@ class ModelProviderCredentials(ProviderCredentials):
     api_version: SecretStr | None = UserConfigurable(default=None)
     deployment_id: SecretStr | None = UserConfigurable(default=None)
 
-    def unmasked(self) -> dict:
-        return unmask(self)
-
     class Config:
         extra = "ignore"
 
 
-def unmask(model: BaseModel):
-    unmasked_fields = {}
-    for field_name, field in model.__fields__.items():
-        value = getattr(model, field_name)
-        if isinstance(value, SecretStr):
-            unmasked_fields[field_name] = value.get_secret_value()
-        else:
-            unmasked_fields[field_name] = value
-    return unmasked_fields
-
-
 class ModelProviderUsage(ProviderUsage):
     """Usage for a particular model from a model provider."""
 

autogpts/autogpt/autogpt/core/resource/schema.py

@@ -1,7 +1,7 @@
 import abc
 import enum
 
-from pydantic import SecretBytes, SecretField, SecretStr
+from pydantic import BaseModel, SecretBytes, SecretField, SecretStr
 
 from autogpt.core.configuration import (
     SystemConfiguration,
@@ -39,6 +39,9 @@ class ProviderBudget(SystemConfiguration):
 class ProviderCredentials(SystemConfiguration):
     """Struct for credentials."""
 
+    def unmasked(self) -> dict:
+        return unmask(self)
+
     class Config:
         json_encoders = {
             SecretStr: lambda v: v.get_secret_value() if v else None,
@@ -47,6 +50,17 @@ class ProviderCredentials(SystemConfiguration):
         }
 
 
+def unmask(model: BaseModel):
+    unmasked_fields = {}
+    for field_name, _ in model.__fields__.items():
+        value = getattr(model, field_name)
+        if isinstance(value, SecretStr):
+            unmasked_fields[field_name] = value.get_secret_value()
+        else:
+            unmasked_fields[field_name] = value
+    return unmasked_fields
+
+
 class ProviderSettings(SystemSettings):
     resource_type: ResourceType
     credentials: ProviderCredentials | None = None

autogpts/autogpt/autogpt/file_workspace/__init__.py

@@ -1,5 +1,46 @@
-from .file_workspace import FileWorkspace
+import enum
+from pathlib import Path
+from typing import Optional
+
+from .base import FileWorkspace
+
+
+class FileWorkspaceBackendName(str, enum.Enum):
+    LOCAL = "local"
+    GCS = "gcs"
+    S3 = "s3"
+
+
+def get_workspace(
+    backend: FileWorkspaceBackendName, *, id: str = "", root_path: Optional[Path] = None
+) -> FileWorkspace:
+    assert bool(root_path) != bool(id), "Specify root_path or id to get workspace"
+    if root_path is None:
+        root_path = Path(f"workspaces/{id}")
+
+    match backend:
+        case FileWorkspaceBackendName.LOCAL:
+            from .local import FileWorkspaceConfiguration, LocalFileWorkspace
+
+            config = FileWorkspaceConfiguration.from_env()
+            config.root = root_path
+            return LocalFileWorkspace(config)
+        case FileWorkspaceBackendName.S3:
+            from .s3 import S3FileWorkspace, S3FileWorkspaceConfiguration
+
+            config = S3FileWorkspaceConfiguration.from_env()
+            config.root = root_path
+            return S3FileWorkspace(config)
+        case FileWorkspaceBackendName.GCS:
+            from .gcs import GCSFileWorkspace, GCSFileWorkspaceConfiguration
+
+            config = GCSFileWorkspaceConfiguration.from_env()
+            config.root = root_path
+            return GCSFileWorkspace(config)
+
 
 __all__ = [
     "FileWorkspace",
+    "FileWorkspaceBackendName",
+    "get_workspace",
 ]

autogpts/autogpt/autogpt/file_workspace/base.py

@@ -3,18 +3,23 @@ The FileWorkspace class provides an interface for interacting with a file worksp
 """
 from __future__ import annotations
 
-import inspect
 import logging
+from abc import ABC, abstractmethod
 from pathlib import Path
-from typing import Any, Callable, Optional
+from typing import Any, Callable, Literal, Optional, overload
+
+from autogpt.core.configuration.schema import SystemConfiguration
 
 logger = logging.getLogger(__name__)
 
 
-class FileWorkspace:
-    """A class that represents a file workspace."""
+class FileWorkspaceConfiguration(SystemConfiguration):
+    restrict_to_root: bool = True
+    root: Path = Path("/")
 
-    NULL_BYTES = ["\0", "\000", "\x00", "\u0000"]
+
+class FileWorkspace(ABC):
+    """A class that represents a file workspace."""
 
     on_write_file: Callable[[Path], Any] | None = None
     """
@@ -24,22 +29,55 @@ class FileWorkspace:
         Path: The path of the file that was written, relative to the workspace root.
     """
 
-    def __init__(self, root: str | Path, restrict_to_root: bool):
-        self._root = self._sanitize_path(root)
-        self._restrict_to_root = restrict_to_root
-
     @property
+    @abstractmethod
     def root(self) -> Path:
-        """The root directory of the file workspace."""
-        return self._root
+        """The root path of the file workspace."""
 
     @property
-    def restrict_to_root(self):
-        """Whether to restrict generated paths to the root."""
-        return self._restrict_to_root
+    @abstractmethod
+    def restrict_to_root(self) -> bool:
+        """Whether to restrict file access to within the workspace's root path."""
 
+    @abstractmethod
     def initialize(self) -> None:
-        self.root.mkdir(exist_ok=True, parents=True)
+        """
+        Calling `initialize()` should bring the workspace to a ready-to-use state.
+        For example, it can create the resource in which files will be stored, if it
+        doesn't exist yet. E.g. a folder on disk, or an S3 Bucket.
+        """
+
+    @abstractmethod
+    def open_file(self, path: str | Path, mode: str = "r"):
+        """Open a file in the workspace."""
+
+    @overload
+    @abstractmethod
+    def read_file(self, path: str | Path, binary: Literal[False] = False) -> str:
+        """Read a file in the workspace as text."""
+        ...
+
+    @overload
+    @abstractmethod
+    def read_file(self, path: str | Path, binary: Literal[True] = True) -> bytes:
+        """Read a file in the workspace as binary."""
+        ...
+
+    @abstractmethod
+    def read_file(self, path: str | Path, binary: bool = False) -> str | bytes:
+        """Read a file in the workspace."""
+
+    @abstractmethod
+    async def write_file(self, path: str | Path, content: str | bytes) -> None:
+        """Write to a file in the workspace."""
+
+    @abstractmethod
+    def list_files(self, path: str | Path = ".") -> list[Path]:
+        """List all files in a directory in the workspace."""
+
+    @abstractmethod
+    def delete_file(self, path: str | Path) -> None:
+        """Delete a file in the workspace."""
 
     def get_path(self, relative_path: str | Path) -> Path:
         """Get the full path for an item in the workspace.
@@ -50,44 +88,7 @@ class FileWorkspace:
         Returns:
             Path: The resolved path relative to the workspace.
         """
-        return self._sanitize_path(
-            relative_path,
-            root=self.root,
-            restrict_to_root=self.restrict_to_root,
-        )
-
-    def open_file(self, path: str | Path, mode: str = "r"):
-        """Open a file in the workspace."""
-        full_path = self.get_path(path)
-        return open(full_path, mode)
-
-    def read_file(self, path: str | Path, binary: bool = False):
-        """Read a file in the workspace."""
-        with self.open_file(path, "rb" if binary else "r") as file:
-            return file.read()
-
-    async def write_file(self, path: str | Path, content: str | bytes):
-        """Write to a file in the workspace."""
-        with self.open_file(path, "wb" if type(content) is bytes else "w") as file:
-            file.write(content)
-
-        if self.on_write_file:
-            path = Path(path)
-            if path.is_absolute():
-                path = path.relative_to(self.root)
-            res = self.on_write_file(path)
-            if inspect.isawaitable(res):
-                await res
-
-    def list_files(self, path: str | Path = "."):
-        """List all files in a directory in the workspace."""
-        full_path = self.get_path(path)
-        return [str(file) for file in full_path.glob("*") if file.is_file()]
-
-    def delete_file(self, path: str | Path):
-        """Delete a file in the workspace."""
-        full_path = self.get_path(path)
-        full_path.unlink()
+        return self._sanitize_path(relative_path, self.root)
 
     @staticmethod
     def _sanitize_path(
@@ -113,9 +114,8 @@ class FileWorkspace:
         # Posix systems disallow null bytes in paths. Windows is agnostic about it.
         # Do an explicit check here for all sorts of null byte representations.
 
-        for null_byte in FileWorkspace.NULL_BYTES:
-            if null_byte in str(relative_path) or null_byte in str(root):
-                raise ValueError("embedded null byte")
+        if "\0" in str(relative_path) or "\0" in str(root):
+            raise ValueError("embedded null byte")
 
         if root is None:
             return Path(relative_path).resolve()

autogpts/autogpt/autogpt/file_workspace/gcs.py

@@ -0,0 +1,91 @@
+"""
+The GCSWorkspace class provides an interface for interacting with a file workspace, and
+stores the files in a Google Cloud Storage bucket.
+"""
+from __future__ import annotations
+
+import inspect
+import logging
+from pathlib import Path
+
+from google.cloud import storage
+
+from autogpt.core.configuration.schema import UserConfigurable
+
+from .base import FileWorkspace, FileWorkspaceConfiguration
+
+logger = logging.getLogger(__name__)
+
+
+class GCSFileWorkspaceConfiguration(FileWorkspaceConfiguration):
+    bucket: str = UserConfigurable("autogpt", from_env="WORKSPACE_STORAGE_BUCKET")
+
+
+class GCSFileWorkspace(FileWorkspace):
+    """A class that represents a Google Cloud Storage workspace."""
+
+    _bucket: storage.Bucket
+
+    def __init__(self, config: GCSFileWorkspaceConfiguration):
+        self._bucket_name = config.bucket
+        self._root = config.root
+
+        self._gcs = storage.Client()
+        super().__init__()
+
+    @property
+    def root(self) -> Path:
+        """The root directory of the file workspace."""
+        return self._root
+
+    @property
+    def restrict_to_root(self):
+        """Whether to restrict generated paths to the root."""
+        return True
+
+    def initialize(self) -> None:
+        self._bucket = self._gcs.get_bucket(self._bucket_name)
+
+    def get_path(self, relative_path: str | Path) -> Path:
+        return super().get_path(relative_path).relative_to(Path("/"))
+
+    def open_file(self, path: str | Path, mode: str = "r"):
+        """Open a file in the workspace."""
+        path = self.get_path(path)
+        blob = self._bucket.blob(str(path))
+        return blob
+
+    def read_file(self, path: str | Path, binary: bool = False) -> str | bytes:
+        """Read a file in the workspace."""
+        blob = self.open_file(path, "r")
+        file_content = (
+            blob.download_as_text() if not binary else blob.download_as_bytes()
+        )
+        return file_content
+
+    async def write_file(self, path: str | Path, content: str | bytes):
+        """Write to a file in the workspace."""
+        blob = self.open_file(path, "w")
+        blob.upload_from_string(content) if isinstance(
+            content, str
+        ) else blob.upload_from_file(content)
+
+        if self.on_write_file:
+            path = Path(path)
+            if path.is_absolute():
+                path = path.relative_to(self.root)
+            res = self.on_write_file(path)
+            if inspect.isawaitable(res):
+                await res
+
+    def list_files(self, path: str | Path = ".") -> list[Path]:
+        """List all files in a directory in the workspace."""
+        path = self.get_path(path)
+        blobs = self._bucket.list_blobs(prefix=str(path))
+        return [Path(blob.name) for blob in blobs if not blob.name.endswith("/")]
+
+    def delete_file(self, path: str | Path) -> None:
+        """Delete a file in the workspace."""
+        path = self.get_path(path)
+        blob = self._bucket.blob(str(path))
+        blob.delete()

autogpts/autogpt/autogpt/file_workspace/local.py

@@ -0,0 +1,67 @@
+"""
+The LocalFileWorkspace class implements a FileWorkspace that works with local files.
+"""
+from __future__ import annotations
+
+import inspect
+import logging
+from pathlib import Path
+
+from .base import FileWorkspace, FileWorkspaceConfiguration
+
+logger = logging.getLogger(__name__)
+
+
+class LocalFileWorkspace(FileWorkspace):
+    """A class that represents a file workspace."""
+
+    def __init__(self, config: FileWorkspaceConfiguration):
+        self._root = self._sanitize_path(config.root)
+        self._restrict_to_root = config.restrict_to_root
+        super().__init__()
+
+    @property
+    def root(self) -> Path:
+        """The root directory of the file workspace."""
+        return self._root
+
+    @property
+    def restrict_to_root(self):
+        """Whether to restrict generated paths to the root."""
+        return self._restrict_to_root
+
+    def initialize(self) -> None:
+        self.root.mkdir(exist_ok=True, parents=True)
+
+    def open_file(self, path: str | Path, mode: str = "r"):
+        """Open a file in the workspace."""
+        full_path = self.get_path(path)
+        return open(full_path, mode)
+
+    def read_file(self, path: str | Path, binary: bool = False):
+        """Read a file in the workspace."""
+        with self.open_file(path, "rb" if binary else "r") as file:
+            return file.read()
+
+    async def write_file(self, path: str | Path, content: str | bytes):
+        """Write to a file in the workspace."""
+        with self.open_file(path, "wb" if type(content) is bytes else "w") as file:
+            file.write(content)
+
+        if self.on_write_file:
+            path = Path(path)
+            if path.is_absolute():
+                path = path.relative_to(self.root)
+            res = self.on_write_file(path)
+            if inspect.isawaitable(res):
+                await res
+
+    def list_files(self, path: str | Path = "."):
+        """List all files in a directory in the workspace."""
+        full_path = self.get_path(path)
+        return [str(file) for file in full_path.glob("*") if file.is_file()]
+
+    def delete_file(self, path: str | Path):
+        """Delete a file in the workspace."""
+        full_path = self.get_path(path)
+        full_path.unlink()

autogpts/autogpt/autogpt/file_workspace/s3.py

@@ -0,0 +1,122 @@
+"""
+The S3Workspace class provides an interface for interacting with a file workspace, and
+stores the files in an S3 bucket.
+"""
+from __future__ import annotations
+
+import contextlib
+import inspect
+import logging
+import os
+from pathlib import Path
+from typing import TYPE_CHECKING, Optional
+
+import boto3
+import botocore.exceptions
+from pydantic import SecretStr
+
+from autogpt.core.configuration.schema import UserConfigurable
+
+from .base import FileWorkspace, FileWorkspaceConfiguration
+
+if TYPE_CHECKING:
+    import mypy_boto3_s3
+
+logger = logging.getLogger(__name__)
+
+
+class S3FileWorkspaceConfiguration(FileWorkspaceConfiguration):
+    bucket: str = UserConfigurable("autogpt", from_env="WORKSPACE_STORAGE_BUCKET")
+    s3_endpoint_url: Optional[SecretStr] = UserConfigurable(
+        from_env=lambda: SecretStr(v) if (v := os.getenv("S3_ENDPOINT_URL")) else None
+    )
+
+
+class S3FileWorkspace(FileWorkspace):
+    """A class that represents an S3 workspace."""
+
+    _bucket: mypy_boto3_s3.service_resource.Bucket
+
+    def __init__(self, config: S3FileWorkspaceConfiguration):
+        self._bucket_name = config.bucket
+        self._root = config.root
+
+        # https://boto3.amazonaws.com/v1/documentation/api/latest/guide/configuration.html
+        self._s3 = boto3.resource(
+            "s3",
+            endpoint_url=config.s3_endpoint_url.get_secret_value()
+            if config.s3_endpoint_url
+            else None,
+        )
+
+        super().__init__()
+
+    @property
+    def root(self) -> Path:
+        """The root directory of the file workspace."""
+        return self._root
+
+    @property
+    def restrict_to_root(self):
+        """Whether to restrict generated paths to the root."""
+        return True
+
+    def initialize(self) -> None:
+        try:
+            self._s3.meta.client.head_bucket(Bucket=self._bucket_name)
+            self._bucket = self._s3.Bucket(self._bucket_name)
+        except botocore.exceptions.ClientError as e:
+            if "(404)" not in str(e):
+                raise
+            self._bucket = self._s3.create_bucket(Bucket=self._bucket_name)
+
+    def get_path(self, relative_path: str | Path) -> Path:
+        return super().get_path(relative_path).relative_to(Path("/"))
+
+    def open_file(self, path: str | Path, mode: str = "r"):
+        """Open a file in the workspace."""
+        path = self.get_path(path)
+        obj = self._bucket.Object(str(path))
+        with contextlib.suppress(botocore.exceptions.ClientError):
+            obj.load()
+        return obj
+
+    def read_file(self, path: str | Path, binary: bool = False) -> str | bytes:
+        """Read a file in the workspace."""
+        file_content = self.open_file(path, "r").get()["Body"].read()
+        return file_content if binary else file_content.decode()
+
+    async def write_file(self, path: str | Path, content: str | bytes):
+        """Write to a file in the workspace."""
+        obj = self.open_file(path, "w")
+        obj.put(Body=content)
+
+        if self.on_write_file:
+            path = Path(path)
+            if path.is_absolute():
+                path = path.relative_to(self.root)
+            res = self.on_write_file(path)
+            if inspect.isawaitable(res):
+                await res
+
+    def list_files(self, path: str | Path = ".") -> list[Path]:
+        """List all files in a directory in the workspace."""
+        path = self.get_path(path)
+        if path == Path("."):
+            return [
+                Path(obj.key)
+                for obj in self._bucket.objects.all()
+                if not obj.key.endswith("/")
+            ]
+        else:
+            return [
+                Path(obj.key)
+                for obj in self._bucket.objects.filter(Prefix=str(path))
+                if not obj.key.endswith("/")
+            ]
+
+    def delete_file(self, path: str | Path) -> None:
+        """Delete a file in the workspace."""
+        path = self.get_path(path)
+        obj = self._s3.Object(self._bucket_name, str(path))
+        obj.delete()

autogpts/autogpt/docker-compose.yml

@@ -16,3 +16,34 @@ services:
       - ./docker-compose.yml:/app/docker-compose.yml:ro
       - ./Dockerfile:/app/Dockerfile:ro
     profiles: ["exclude-from-up"]
+
+  # Only for TESTING purposes. Run with: docker compose run --build --rm autogpt-test
+  autogpt-test:
+    build: ./
+    env_file:
+      - .env
+    environment:
+      S3_ENDPOINT_URL: http://minio:9000
+      AWS_ACCESS_KEY_ID: minio
+      AWS_SECRET_ACCESS_KEY: minio123
+    entrypoint: ["poetry", "run"]
+    command: ["pytest", "-v"]
+    volumes:
+      - ./autogpt:/app/autogpt
+      - ./tests:/app/tests
+    depends_on:
+      - minio
+    profiles: ["exclude-from-up"]
+  minio:
+    image: minio/minio
+    environment:
+      MINIO_ACCESS_KEY: minio
+      MINIO_SECRET_KEY: minio123
+    ports:
+      - 9000:9000
+    volumes:
+      - minio-data:/data
+    command: server /data
+    profiles: ["exclude-from-up"]
+volumes:
+  minio-data:

autogpts/autogpt/pyproject.toml

@@ -25,6 +25,7 @@ python = "^3.10"
 # autogpt-forge = { path = "../forge" }
 autogpt-forge = {git = "https://github.com/Significant-Gravitas/AutoGPT.git", subdirectory = "autogpts/forge"}
 beautifulsoup4 = "^4.12.2"
+boto3 = "^1.33.6"
 charset-normalizer = "^3.1.0"
 click = "*"
 colorama = "^0.4.6"
@@ -68,6 +69,7 @@ openapi-python-client = "^0.14.0"
 # agbenchmark = { path = "../../benchmark", optional = true }
 agbenchmark = {git = "https://github.com/Significant-Gravitas/AutoGPT.git", subdirectory = "benchmark", optional = true}
 google-cloud-logging = "^3.8.0"
+google-cloud-storage = "^2.13.0"
 
 [tool.poetry.extras]
 benchmark = ["agbenchmark"]
@@ -75,6 +77,7 @@ benchmark = ["agbenchmark"]
 [tool.poetry.group.dev.dependencies]
 auto-gpt-plugin-template = {git = "https://github.com/Significant-Gravitas/Auto-GPT-Plugin-Template", rev = "0.1.0"}
 black = "*"
+boto3-stubs = {extras = ["s3"], version = "^1.33.6"}
 flake8 = "*"
 gitpython = "^3.1.32"
 isort = "*"

autogpts/autogpt/tests/conftest.py

@@ -11,7 +11,11 @@ from autogpt.agents.agent import Agent, AgentConfiguration, AgentSettings
 from autogpt.app.main import _configure_openai_provider
 from autogpt.config import AIProfile, Config, ConfigBuilder
 from autogpt.core.resource.model_providers import ChatModelProvider, OpenAIProvider
-from autogpt.file_workspace import FileWorkspace
+from autogpt.file_workspace.local import (
+    FileWorkspace,
+    FileWorkspaceConfiguration,
+    LocalFileWorkspace,
+)
 from autogpt.llm.api_manager import ApiManager
 from autogpt.logs.config import configure_logging
 from autogpt.models.command_registry import CommandRegistry
@@ -47,7 +51,7 @@ def workspace_root(agent_data_dir: Path) -> Path:
 
 @pytest.fixture()
 def workspace(workspace_root: Path) -> FileWorkspace:
-    workspace = FileWorkspace(workspace_root, restrict_to_root=True)
+    workspace = LocalFileWorkspace(FileWorkspaceConfiguration(root=workspace_root))
     workspace.initialize()
     return workspace
 

autogpts/autogpt/tests/unit/test_gcs_file_workspace.py

@@ -0,0 +1,108 @@
+import os
+import uuid
+from pathlib import Path
+
+import pytest
+import pytest_asyncio
+from google.cloud.exceptions import NotFound
+
+from autogpt.file_workspace.gcs import GCSFileWorkspace, GCSFileWorkspaceConfiguration
+
+if not os.getenv("GOOGLE_APPLICATION_CREDENTIALS"):
+    pytest.skip("GOOGLE_APPLICATION_CREDENTIALS are not set", allow_module_level=True)
+
+
+@pytest.fixture
+def gcs_bucket_name() -> str:
+    return f"test-bucket-{str(uuid.uuid4())[:8]}"
+
+
+@pytest.fixture
+def gcs_workspace_uninitialized(gcs_bucket_name: str) -> GCSFileWorkspace:
+    os.environ["WORKSPACE_STORAGE_BUCKET"] = gcs_bucket_name
+    ws_config = GCSFileWorkspaceConfiguration.from_env()
+    workspace = GCSFileWorkspace(ws_config)
+    yield workspace  # type: ignore
+    del os.environ["WORKSPACE_STORAGE_BUCKET"]
+
+
+def test_initialize(
+    gcs_bucket_name: str, gcs_workspace_uninitialized: GCSFileWorkspace
+):
+    gcs = gcs_workspace_uninitialized._bucket
+
+    # test that the bucket doesn't exist yet
+    with pytest.raises(NotFound):
+        gcs.get_blob(gcs_bucket_name)
+
+    gcs_workspace_uninitialized.initialize()
+
+    # test that the bucket has been created
+    gcs.get_blob(gcs_bucket_name)
+
+
+def test_workspace_bucket_name(
+    gcs_workspace: GCSFileWorkspace,
+    gcs_bucket_name: str,
+):
+    assert gcs_workspace._bucket.name == gcs_bucket_name
+
+
+@pytest.fixture
+def gcs_workspace(gcs_workspace_uninitialized: GCSFileWorkspace) -> GCSFileWorkspace:
+    (gcs_workspace := gcs_workspace_uninitialized).initialize()
+    yield gcs_workspace  # type: ignore
+
+    # Empty & delete the test bucket
+    gcs_workspace._bucket.delete_blobs(gcs_workspace._bucket.list_blobs())
+    gcs_workspace._bucket.delete()
+
+
+TEST_FILES: list[tuple[str | Path, str]] = [
+    ("existing_test_file_1", "test content 1"),
+    ("existing_test_file_2.txt", "test content 2"),
+    (Path("existing_test_file_3"), "test content 3"),
+    (Path("existing/test/file/4"), "test content 4"),
+]
+
+
+@pytest_asyncio.fixture
+async def gcs_workspace_with_files(gcs_workspace: GCSFileWorkspace) -> GCSFileWorkspace:
+    for file_name, file_content in TEST_FILES:
+        gcs_workspace._bucket.blob(str(file_name)).upload_from_string(file_content)
+    yield gcs_workspace  # type: ignore
+
+
+@pytest.mark.asyncio
+async def test_read_file(gcs_workspace_with_files: GCSFileWorkspace):
+    for file_name, file_content in TEST_FILES:
+        content = gcs_workspace_with_files.read_file(file_name)
+        assert content == file_content
+
+    with pytest.raises(NotFound):
+        gcs_workspace_with_files.read_file("non_existent_file")
+
+
+def test_list_files(gcs_workspace_with_files: GCSFileWorkspace):
+    files = gcs_workspace_with_files.list_files()
+    assert set(files) == set(Path(file_name) for file_name, _ in TEST_FILES)
+
+
+@pytest.mark.asyncio
+async def test_write_read_file(gcs_workspace: GCSFileWorkspace):
+    await gcs_workspace.write_file("test_file", "test_content")
+    assert gcs_workspace.read_file("test_file") == "test_content"
+
+
+@pytest.mark.asyncio
+async def test_overwrite_file(gcs_workspace_with_files: GCSFileWorkspace):
+    for file_name, _ in TEST_FILES:
+        await gcs_workspace_with_files.write_file(file_name, "new content")
+        assert gcs_workspace_with_files.read_file(file_name) == "new content"
+
+
+def test_delete_file(gcs_workspace_with_files: GCSFileWorkspace):
+    for file_to_delete, _ in TEST_FILES:
+        gcs_workspace_with_files.delete_file(file_to_delete)
+        with pytest.raises(NotFound):
+            gcs_workspace_with_files.read_file(file_to_delete)

autogpts/autogpt/tests/unit/test_local_file_workspace.py

@@ -1,9 +1,8 @@
-import itertools
 from pathlib import Path
 
 import pytest
 
-from autogpt.file_workspace import FileWorkspace
+from autogpt.file_workspace.local import FileWorkspaceConfiguration, LocalFileWorkspace
 
 _WORKSPACE_ROOT = Path("home/users/monty/auto_gpt_workspace")
 
@@ -31,17 +30,11 @@ _INACCESSIBLE_PATHS = (
         Path("test_folder/../../not_auto_gpt_workspace/test_file.txt"),
     ]
     + [
-        # Contains null bytes
-        Path(template.format(null_byte=null_byte))
-        for template, null_byte in itertools.product(
-            [
-                "{null_byte}",
-                "{null_byte}test_file.txt",
-                "test_folder/{null_byte}",
-                "test_folder/{null_byte}test_file.txt",
-            ],
-            FileWorkspace.NULL_BYTES,
-        )
+        # Contains null byte
+        Path("\0"),
+        Path("\0test_file.txt"),
+        Path("test_folder/\0"),
+        Path("test_folder/\0test_file.txt"),
     ]
     + [
         # Absolute paths
@@ -68,7 +61,7 @@ def inaccessible_path(request):
 
 
 def test_sanitize_path_accessible(accessible_path, workspace_root):
-    full_path = FileWorkspace._sanitize_path(
+    full_path = LocalFileWorkspace._sanitize_path(
         accessible_path,
         root=workspace_root,
         restrict_to_root=True,
@@ -79,7 +72,7 @@ def test_sanitize_path_accessible(accessible_path, workspace_root):
 
 def test_sanitize_path_inaccessible(inaccessible_path, workspace_root):
     with pytest.raises(ValueError):
-        FileWorkspace._sanitize_path(
+        LocalFileWorkspace._sanitize_path(
             inaccessible_path,
             root=workspace_root,
             restrict_to_root=True,
@@ -87,13 +80,13 @@ def test_sanitize_path_inaccessible(inaccessible_path, workspace_root):
 
 
 def test_get_path_accessible(accessible_path, workspace_root):
-    workspace = FileWorkspace(workspace_root, True)
+    workspace = LocalFileWorkspace(FileWorkspaceConfiguration(root=workspace_root))
     full_path = workspace.get_path(accessible_path)
     assert full_path.is_absolute()
     assert full_path.is_relative_to(workspace_root)
 
 
 def test_get_path_inaccessible(inaccessible_path, workspace_root):
-    workspace = FileWorkspace(workspace_root, True)
+    workspace = LocalFileWorkspace(FileWorkspaceConfiguration(root=workspace_root))
     with pytest.raises(ValueError):
         workspace.get_path(inaccessible_path)

autogpts/autogpt/tests/unit/test_s3_file_workspace.py

@@ -0,0 +1,106 @@
+import os
+import uuid
+from pathlib import Path
+
+import pytest
+import pytest_asyncio
+from botocore.exceptions import ClientError
+
+from autogpt.file_workspace.s3 import S3FileWorkspace, S3FileWorkspaceConfiguration
+
+if not os.getenv("S3_ENDPOINT_URL") and not os.getenv("AWS_ACCESS_KEY_ID"):
+    pytest.skip("S3 environment variables are not set", allow_module_level=True)
+
+
+@pytest.fixture
+def s3_bucket_name() -> str:
+    return f"test-bucket-{str(uuid.uuid4())[:8]}"
+
+
+@pytest.fixture
+def s3_workspace_uninitialized(s3_bucket_name: str) -> S3FileWorkspace:
+    os.environ["WORKSPACE_STORAGE_BUCKET"] = s3_bucket_name
+    ws_config = S3FileWorkspaceConfiguration.from_env()
+    workspace = S3FileWorkspace(ws_config)
+    yield workspace  # type: ignore
+    del os.environ["WORKSPACE_STORAGE_BUCKET"]
+
+
+def test_initialize(s3_bucket_name: str, s3_workspace_uninitialized: S3FileWorkspace):
+    s3 = s3_workspace_uninitialized._s3
+
+    # test that the bucket doesn't exist yet
+    with pytest.raises(ClientError):
+        s3.meta.client.head_bucket(Bucket=s3_bucket_name)
+
+    s3_workspace_uninitialized.initialize()
+
+    # test that the bucket has been created
+    s3.meta.client.head_bucket(Bucket=s3_bucket_name)
+
+
+def test_workspace_bucket_name(
+    s3_workspace: S3FileWorkspace,
+    s3_bucket_name: str,
+):
+    assert s3_workspace._bucket.name == s3_bucket_name
+
+
+@pytest.fixture
+def s3_workspace(s3_workspace_uninitialized: S3FileWorkspace) -> S3FileWorkspace:
+    (s3_workspace := s3_workspace_uninitialized).initialize()
+    yield s3_workspace  # type: ignore
+
+    # Empty & delete the test bucket
+    s3_workspace._bucket.objects.all().delete()
+    s3_workspace._bucket.delete()
+
+
+TEST_FILES: list[tuple[str | Path, str]] = [
+    ("existing_test_file_1", "test content 1"),
+    ("existing_test_file_2.txt", "test content 2"),
+    (Path("existing_test_file_3"), "test content 3"),
+    (Path("existing/test/file/4"), "test content 4"),
+]
+
+
+@pytest_asyncio.fixture
+async def s3_workspace_with_files(s3_workspace: S3FileWorkspace) -> S3FileWorkspace:
+    for file_name, file_content in TEST_FILES:
+        s3_workspace._bucket.Object(str(file_name)).put(Body=file_content)
+    yield s3_workspace  # type: ignore
+
+
+@pytest.mark.asyncio
+async def test_read_file(s3_workspace_with_files: S3FileWorkspace):
+    for file_name, file_content in TEST_FILES:
+        content = s3_workspace_with_files.read_file(file_name)
+        assert content == file_content
+
+    with pytest.raises(ClientError):
+        s3_workspace_with_files.read_file("non_existent_file")
+
+
+def test_list_files(s3_workspace_with_files: S3FileWorkspace):
+    files = s3_workspace_with_files.list_files()
+    assert set(files) == set(Path(file_name) for file_name, _ in TEST_FILES)
+
+
+@pytest.mark.asyncio
+async def test_write_read_file(s3_workspace: S3FileWorkspace):
+    await s3_workspace.write_file("test_file", "test_content")
+    assert s3_workspace.read_file("test_file") == "test_content"
+
+
+@pytest.mark.asyncio
+async def test_overwrite_file(s3_workspace_with_files: S3FileWorkspace):
+    for file_name, _ in TEST_FILES:
+        await s3_workspace_with_files.write_file(file_name, "new content")
+        assert s3_workspace_with_files.read_file(file_name) == "new content"
+
+
+def test_delete_file(s3_workspace_with_files: S3FileWorkspace):
+    for file_to_delete, _ in TEST_FILES:
+        s3_workspace_with_files.delete_file(file_to_delete)
+        with pytest.raises(ClientError):
+            s3_workspace_with_files.read_file(file_to_delete)