ACA-Py provides an OpenAPI-documented REST interface for administering the agent's internal state and initiating communication with connected agents.
To see the specifics of the supported endpoints, as well as the expected request and response formats, it is recommended to run the aca-py
agent with the --admin {HOST} {PORT}
and --admin-insecure-mode
command line parameters. This exposes the OpenAPI UI on the provided port for interaction via a web browser. For production deployments, run the agent with --admin-api-key {KEY}
and add the X-API-Key: {KEY}
header to all requests instead of using the --admin-insecure-mode
parameter.
To invoke a specific method:
- Scroll to and find that endpoint;
- Click on the endpoint name to expand its section of the UI;
- Click on the Try it out button;
- Fill in any data necessary to run the command;
- Click Execute;
- Check the response to see if the request worked as expected.
The mechanical steps are easy; however, the fourth step from the list above can be tricky. Supplying the right data and, where JSON is involved, getting the syntax correct—braces and quotes can be a pain. When steps don't work, start your debugging by looking at your JSON. You may also choose to use a REST client like Postman or Insomnia, which will provide syntax highlighting and other features to simplify the process.
Because API methods often initiate asynchronous processes, the JSON response provided by an endpoint is not always sufficient to determine the next action. To handle this situation, as well as events triggered by external inputs (such as new connection requests), it is necessary to implement a webhook processor, as detailed in the next section.
The combination of an OpenAPI client and webhook processor is referred to as an ACA-Py Controller and is the recommended method to define custom behaviors for your ACA-Py-based agent application.
When ACA-Py is started with the --webhook-url {URL}
command line parameter, state-management records are sent to the provided URL via POST requests whenever a record is created or its state
property is updated.
When a webhook is dispatched, the record topic
is appended as a path component to the URL. For example, https://webhook.host.example
becomes https://webhook.host.example/topic/connections
when a connection record is updated. A POST request is made to the resulting URL with the body of the request comprising a serialized JSON object. The full set of properties of the current set of webhook payloads are listed below. Note that empty (null-value) properties are omitted.
ACA-Py's Admin API also supports delivering webhooks over WebSocket. This can be especially useful when working with scripts that interact with the Admin API but don't have a web server listening to recieve webhooks in response to its actions. No additional command line parameters are required to enable WebSocket support.
Webhooks received over WebSocket will contain the same data as webhooks posted over http but the structure differs in order to communicate details that would have been received as part of the HTTP request path and headers.
topic
: The topic of the webhook, such asconnections
orbasicmessages
payload
: The payload of the webhook; this is the data usually received in the request body when webhooks are delivered over HTTPwallet_id
: If using multitenancy, this is the wallet ID of the subwallet that emitted the webhook. This value will be omitted if not using multitenancy.
To open a WebSocket, connect to the /ws
endpoint of the Admin API.
connection_id
: the unique connection identifierstate
:init
/invitation
/request
/response
/active
/error
/inactive
my_did
: the DID this agent is using in the connectiontheir_did
: the DID the other agent in the connection is usingtheir_label
: a connection label provided by the other agenttheir_role
: a role assigned to the other agent in the connectioninbound_connection_id
: a connection identifier for the related inbound routing connectioninitiator
:self
/external
/multiuse
invitation_key
: a verification key used to identify the source connection invitationrequest_id
: the@id
property from the connection request messagerouting_state
:none
/request
/active
/error
accept
:manual
/auto
error_msg
: the most recent error messageinvitation_mode
:once
/multi
alias
: a local alias for the connection record
connection_id
: the identifier of the related pairwise connectionmessage_id
: the@id
of the incoming agent messagecontent
: the contents of the agent messagestate
:received
Enable using --monitor-forward
.
connection_id
: the identifier of the connection associated with the recipient keyrecipient_key
: the recipient key of the forward message (to
field of the forward message)status
: The delivery status of the received forward message. Possible values:sent_to_session
: Message is sent directly to the connection over an active transport sessionsent_to_external_queue
: Message is sent to an external queue. No information is known on the delivery of the messagequeued_for_delivery
: Message is queued for delivery using outbound transport (recipient connection has an endpoint)waiting_for_pickup
: The connection has no reachable endpoint. Need to wait for the recipient to connect with return routing for deliveryundeliverable
: The connection has no reachable endpoint, and the internal queue for messages is not enabled (--enable-undelivered-queue
).
credential_exchange_id
: the unique identifier of the credential exchangeconnection_id
: the identifier of the related pairwise connectionthread_id
: the thread ID of the previously received credential proposal or offerparent_thread_id
: the parent thread ID of the previously received credential proposal or offerinitiator
: issue-credential exchange initiatorself
/external
state
:proposal_sent
/proposal_received
/offer_sent
/offer_received
/request_sent
/request_received
/issued
/credential_received
/credential_acked
credential_definition_id
: the ledger identifier of the related credential definitionschema_id
: the ledger identifier of the related credential schemacredential_proposal_dict
: the credential proposal messagecredential_offer
: (Indy) credential offercredential_request
: (Indy) credential requestcredential_request_metadata
: (Indy) credential request metadatacredential_id
: the wallet identifier of the stored credentialraw_credential
: the credential record as receivedcredential
: the credential record as stored in the walletauto_offer
: (boolean) whether to automatically offer the credentialauto_issue
: (boolean) whether to automatically issue the credentialerror_msg
: the previous error message
presentation_exchange_id
: the unique identifier of the presentation exchangeconnection_id
: the identifier of the related pairwise connectionthread_id
: the thread ID of the previously received presentation proposal or offerinitiator
: present-proof exchange initiator:self
/external
state
:proposal_sent
/proposal_received
/request_sent
/request_received
/presentation_sent
/presentation_received
/verified
presentation_proposal_dict
: the presentation proposal messagepresentation_request
: (Indy) presentation request (also known as proof request)presentation
: (Indy) presentation (also known as proof)verified
: (string) whether the presentation is verified:true
orfalse
auto_present
: (boolean) prover choice to auto-present proof as verifier requestserror_msg
: the previous error message
The best way to develop a new admin API or protocol is to follow one of the existing protocols, such as the Credential Exchange or Presentation Exchange.
The routes.py
file contains the API definitions - API endpoints and payload schemas (note that these are not the Aries message schemas).
The payload schemas are defined using marshmallow and will be validated automatically when the API is executed (using middleware). (This raises a status 422
HTTP response with an error message if the schema validation fails.)
API endpoints are defined using aiohttp_apispec tags (e.g. @doc
, @request_schema
, @response_schema
etc.) which define the input and output parameters of the endpoint. API URL paths are defined in the register()
method and added to the Swagger page in the post_process_routes()
method.
The APIs should return the following HTTP status:
- HTTP 200 for successful API completion, with an appropriate response
- HTTP 400 (or appropriate 4xx code) (with an error message) for errors on input parameters (i.e., the user can retry with different parameters and potentially get a successful API call)
- HTTP 404 if a record is expected and not found (generally for GET requests that fetch a single record)
- HTTP 500 (or appropriate 5xx code) if there is some other processing error (i.e., it won't make any difference what parameters the user tries) with an error message
...and should not return:
- HTTP 500 with a stack trace due to an untrapped error (we should handle error conditions with a 400 or 404 response and catch errors, providing a meaningful error message)