Self-improving AI Employees for everyone
AWS Service
Self-contained plugin at server/nodes/aws/ over the
official AWS SDK for Python (boto3, a server dependency). One
dual-purpose node, awsAction (workflow node + AI tool aws_action): typed
operations for the common flows (identity, EC2 instances, S3 objects), call
for any operation of any AWS service, and list_operations to discover them.
Why the SDK and not the AWS CLI (unlike the gcloud / cf / gh plugins): AWS
CLI v2 ships as an MSI on Windows and a .pkg on macOS, with no portable
archive to pin and extract the way gcloud/_install.py does, and its browser
login (aws login) cannot run on a remote or hosted server. The SDK installs
with uv sync on every platform, and a stored key pair works headless.
Auth model: an IAM access key pair in OpenCompany's encrypted store.
AwsCredential (ApiKeyCredential, id aws) keeps three rows: the Secret
Access Key under the provider id (the field the credentials modal validates),
aws_access_key_id, and the optional aws_region (extra_fields, the
cloudflare companion-field pattern). The modal probe signs
sts:GetCallerIdentity with the pair; every IAM identity may call it, so a
least-privilege key validates too. The Access Key ID must be saved before the
secret is validated, and the probe says so when it is missing.
No ambient credentials. Every client is created with the key pair passed
explicitly, which outranks the SDK's credential chain, so environment
variables, ~/.aws and an EC2 instance profile on the server never
authenticate a node.
File map
| File | Role |
|---|---|
__init__.py |
register_output_schema("awsAction", …) + register_option_loader("awsServices" / "awsOperations", …); the node auto-registers on import |
aws_action.py |
AwsActionNode(ActionNode), usable_as_tool=True, tool_name="aws_action", both canvas handles kept visible, eleven @Operations; _keys resolves the credential before any work, so a missing key gives the PermissionDeniedError envelope; _shape drops only None |
_sdk.py |
Every boto3 / botocore touch, imported lazily: introspection (service_names, operations, resolve_operation with difflib suggestions), calls (call, collect over a paginator with a JMESPath projection, download_object), to_plain, raise_user_error, and the two dropdown loaders |
_credentials.py |
AwsCredential: resolve() also requires the Access Key ID; _probe() reads the companion rows and calls STS |
icon.svg / aws.svg / meta.json |
Node icon and co-located credential icon (lobehub's AWS mark on a dark tile, readable through <img> on light and dark themes); colour #FF9900 |
Paired skill: server/skills/aws/aws-skill/SKILL.md
(bound in visuals.json: "awsAction": {"skill": "aws-skill"}). Palette
group: deployment. Catalogue entry: kind: "apiKey" with the three fields
in credential_providers.json.
Clients
One botocore session (module-level, cached) holds the service models and
client classes for every call; it holds no credentials. Each operation creates
a client with the key pair and region passed explicitly, about 5 ms once the
service's model is loaded (the first client of a service pays ~0.3 s). Client
creation is serialised by a lock because botocore sessions are not
thread-safe; the API calls run in worker threads (asyncio.to_thread)
concurrently. Every client uses 10 s connect / 60 s read timeouts and the
SDK's standard retry mode (3 attempts, throttling and transient errors only).
Operations
| Op | SDK call | Notes |
|---|---|---|
whoami |
sts.get_caller_identity |
|
ec2_instances_list |
ec2.describe_instances paginator, projected per instance (InstanceId, Name tag, State, type, IPs, zone, launch time, image, key) |
instance_state filter; stops at limit, reports truncated |
ec2_instance_describe |
ec2.describe_instances(InstanceIds=[id]) |
The full instance |
ec2_instance_start / _stop |
ec2.start_instances / stop_instances |
|
s3_list |
s3.list_buckets, or the list_objects_v2 paginator (Key, Size, LastModified, StorageClass) |
prefix; stops at limit |
s3_upload |
s3.put_object |
file read by coerce_file_param (workspace containment, 25 MB); key defaults to the file name; content type from the name |
s3_download |
s3.get_object |
Refused past MEDIA_MAX_READ_BYTES before reading; saved with write_media, returned as a FileRef under file |
s3_delete |
s3.delete_object |
|
call |
<service>.<api_operation>(**request) |
Resolved before any credential use; botocore validates request against the model before sending; one page per call |
list_operations |
botocore's bundled models | No credentials, no network |
Responses are shaped the way the AWS CLI prints them: ResponseMetadata
dropped, timestamps as ISO 8601, blobs as base64, a streaming body (Lambda
Payload) read inline up to 1 MB. Errors become NodeUserError with AWS's
code, message and request id, plus a hint for key (InvalidClientTokenId,
SignatureDoesNotMatch, ...), permission (AccessDenied,
UnauthorizedOperation) and throttling codes.
Tests
server/tests/nodes/test_aws.py replaces
_sdk._client with real botocore clients wrapped in botocore.stub.Stubber,
so requests are validated against the real API models without network:
registration and flat tool schema, region precedence, missing-credential
envelopes, AWS error mapping, the EC2 projection and truncation, S3 upload
containment and download to a FileRef, call resolution before
credentials, parameter validation, inline streaming bodies, list_operations
without credentials, the credential probe, a clean-interpreter check that
importing the plugin never loads boto3, and the catalogue / skill / asset
wiring. aws is also in _MIGRATED_PLUGINS of
tests/test_plugin_self_containment.py, and the tool name is pinned in
tests/fixtures/tool_names_snapshot.json.