Supported methods

Summary

Method	Description	Reference	Support
			Python	Go	Java
Session management
create_session	Deprecated, use async_create_session instead		✅	❌	❌
async_create_session			✅	✅	❌
async_get_session			✅	✅	❌
async_list_sessions			✅	✅	❌
async_delete_session			✅	✅	❌
get_session			✅	✅	❌
list_sessions			✅	✅	❌
delete_session			✅	✅	❌
async_get_session			✅	✅	❌
async_list_sessions			✅	✅	❌
async_create_session			✅	✅	❌
async_delete_session			✅	✅	❌
Memory
async_add_session_to_memory			✅	✅	❌
async_search_memory			✅	✅	❌
Query
stream_query			✅	✅	❌
async_stream_query			✅	✅	❌
streaming_agent_run_with_events			✅	✅	❌
Others
register_feedback			✅	✅	❌

AIPlatform API for Agent Engine

Calling the Agent Engine methods

AIPlatform AgentEngine Input

Input is JSON formatted according to the following schema.

{
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "$id": "https://google.com/agentengine/aiplatform.json",
    "title": "AgentEngine wrapper for AIPlatform",
    "description": "",
    "type": "object",
    "properties": {
        "class_method": {
            "description": "Name of the underlying method",
            "type": "string"
        },
        "input": {
            "description": "Method-specific input",
            "type": "object"
        },
    },
    "required": [
        "class_method",
        "input"
    ]
}

Response for streaming

For early failures, before the streaming starts, the response is just like the one for non-streaming version

After the streaming starts, it provides a stream of JSON-formatted data one per line. JSON object is specific for the called method. In case of an error, the HTTP status is already 200 and cannot be changed, so the error is formatted as an streaming data as well, according to the schema:

{
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "$id": "https://google.com/agentengine/streaming_error.json",
    "title": "Defines how streaming errors are formatted",
    "description": "",
    "type": "object",
    "properties": {
        "error": {
            "description": "Object describing the error",
            "type": "object"
        },
    },
    "required": [
        "error"
    ]
}

Response for non-streaming

On success, the response is JSON-formatted response specific for the called method (class_method).

On failure, the response is JSON-formatted error similar to the following one:

{
  "error": {
    "code": 400,
    "message": "Reasoning Engine Execution failed.\nPlease refer to our documentation (https://cloud.google.com/vertex-ai/generative-ai/docs/agent-engine/troubleshooting/use) for checking logs and other troubleshooting tips.\nError Details: {\"detail\":\"Agent Engine Error: An error occurred during invocation. Exception: 404 NOT_FOUND. {'error': {'code': 404, 'message': 'Session projects/kdroste-adk-2025-12/locations/us-central1/reasoningEngines/3765669545214214144/sessions/7177615616973471745 not found.', 'status': 'NOT_FOUND'}}\\nRequest Data: {'session_id': '7177615616973471745', 'user_id': 'u_12345_non_existing'}\"}",
    "status": "FAILED_PRECONDITION",
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.DebugInfo",
        "detail": "[ORIGINAL ERROR] generic::failed_precondition: Reasoning Engine Execution failed.\nPlease refer to our documentation (https://cloud.google.com/vertex-ai/generative-ai/docs/agent-engine/troubleshooting/use) for checking logs and other troubleshooting tips.\nError Details: {\"detail\":\"Agent Engine Error: An error occurred during invocation. Exception: 404 NOT_FOUND. {'error': {'code': 404, 'message': 'Session projects/kdroste-adk-2025-12/locations/us-central1/reasoningEngines/3765669545214214144/sessions/7177615616973471745 not found.', 'status': 'NOT_FOUND'}}\\nRequest Data: {'session_id': '7177615616973471745', 'user_id': 'u_12345_non_existing'}\"} [google.rpc.error_details_ext] { message: \"Reasoning Engine Execution failed.\\nPlease refer to our documentation (https://cloud.google.com/vertex-ai/generative-ai/docs/agent-engine/troubleshooting/use) for checking logs and other troubleshooting tips.\\nError Details: {\\\"detail\\\":\\\"Agent Engine Error: An error occurred during invocation. Exception: 404 NOT_FOUND. {\\'error\\': {\\'code\\': 404, \\'message\\': \\'Session projects/kdroste-adk-2025-12/locations/us-central1/reasoningEngines/3765669545214214144/sessions/7177615616973471745 not found.\\', \\'status\\': \\'NOT_FOUND\\'}}\\\\nRequest Data: {\\'session_id\\': \\'7177615616973471745\\', \\'user_id\\': \\'u_12345_non_existing\\'}\\\"}\" }"
      }
    ]
  }
}

Create session

You should use async_create_session. create_session is deprecated.

async_create_session

Input

Input adheres to the general schema :

{
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "$id": "https://google.com/agentengine/async_create_session.schema.json",
    "title": "Request for async_create_session",
    "description": "",
    "type": "object",
    "properties": {
        "user_id": {
            "description": "UserID",
            "type": "string"
        },
        "state": {
            "description": "Dictionary mapping string to objects",
            "type": "object"
        },
    },
    "required": [
        "user_id"
    ]
}

Semantics

Creates a new session

Output

Session data is returned

TODO(kdroste): create a schema for that

type SessionData struct {
    UserID         string          `json:"user_id"`
    LastUpdateTime float64         `json:"last_update_time"`
    AppName        string          `json:"app_name"`
    ID             string          `json:"id"`
    State          map[string]any  `json:"state"`
    Events         []session.Event `json:"events"`
}

Error handling

In case of a failure and error is returned using the non-streaming version of response

Invocation examples

curl -v \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json" \
https://${LOCATION_ID}-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/${LOCATION_ID}/reasoningEngines/${RESOURCE_ID}:query -d '{
    "class_method": "async_create_session",
    "input": {
        "user_id": "u_12345",
    }
}'