Logging of AI requests

If you need a detailed retrospective log of communications between AI Enterprise and external AI services (such as OpenAI Platform, Azure OpenAI, Google Vertex AI, and so on), IDE Services allows you to track all request-response interactions. The log data can be stored in your organization's object storage or a dedicated object storage for logs.

Additionally, you can configure the formatting style for AI logs, choosing from JSON, HTTP, or Splunk. Examples of different formatting styles:

                {
                    "correlation": "a56996318975f2f8",
                    "protocol": "HTTP/1.1",
                    "method": "GET",
                    "path": "/api/ai/user/v5/llm/profiles",
                    "attributes": {
                        "userId": "5caeeae2-eb90-441d-9cfb-67bc9ec48d3c"
                    },
                    "headers": {
                        "Authorization": [
                            "XXX"
                        ]
                    },
                    "body": "<body>"
                }
            

                Incoming Request: 96cd14698f9ff001
GET https://<serverURL>/api/ai/user/v5/llm/profiles HTTP/1.1
Authorization: XXX
Request Attribute `userId`: 6092281d-947a-4cf4-bb76-70332c94b595

<body>
            

                origin=remote type=request correlation=baa0fe4c12598729 protocol=HTTP/1.1 method=GET path=/api/ai/user/v5/llm/profiles attributes={userId=dca1270f-1e6d-4518-a49d-b48cf38ad7f0} headers={Authorization=[XXX]} body={<body>}
            

Logging configuration

You can configure IDE Services to store logs with AI requests in your existing object storage connected to IDE Services, or in a separate bucket.

Store AI logs in the object storage used by IDE Services

You can choose to use the same object storage as configured for your IDE Services. For this purpose, you need to provide a specific path-prefix value.

Add the following configuration to your server configuration file:
tbe: ai: platform: logging: enabled: true path-prefix: <PATH_PREFIX> format: <json|http|splunk>
ides: configCustomization: ai: platform: logging: enabled: true path-prefix: <PATH_PREFIX> format: <json|http|splunk>

Store AI logs in a separate object storage

If you'd like to store AI requests in a separate object storage, you need to create a dedicated S3 bucket or Azure Storage account and specify its details in the server configuration file.

Add the following configuration to your server configuration file:
tbe: ai: platform: logging: enabled: true format: <json|http|splunk> storage-type: <azure|s3> azure: container: <CONTAINER_NAME> connection-string: <CONNECTION_STRING> s3: bucket: <BUCKET_NAME> url: <ENDPOINT_URL> access-key: <ACCESS_KEY> secret-key: <ACCESS_SECRET>
ides: configCustomization: ai: platform: logging: enabled: true format: <json|http|splunk> storage-type: <azure|s3> azure: container: <CONTAINER_NAME> connection-string: <CONNECTION_STRING> s3: bucket: <BUCKET_NAME> url: <ENDPOINT_URL> access-key: <ACCESS_KEY> secret-key: <ACCESS_SECRET>

ai.platform.logging.enabled: Enables logging of AI requests.
ai.platform.logging.format: Specify one of Logbook's formatting styles for logs. Possible values: json, http, and splunk.
ai.platform.logging.path-prefix: To store AI logs together with other IDE Services Server data, specify the path for these log files in your object storage. The resulting layout is as follows: $pathPrefix/${yyyy-MM-dd}/$correlationId-response.json.
ai.platform.logging.storage-type: If you use a separate object storage for logs, specify its type: s3 or azure. Depending on the specified type, set up a connection to the storage of your choice.
ai.platform.logging.azure.connection-string: Provide a connection string to authorize requests to Azure storage.
ai.platform.logging.azure.container: Specify the name of the Azure container to connect.
ai.platform.logging.s3.url: Specify a URL to access your S3 bucket.
ai.platform.logging.s3.bucket: Specify the name of your S3 bucket.
ai.platform.logging.s3.access-key: Provide a unique key to access the S3-compatible storage.
ai.platform.logging.s3.secretKey: Provide a secret key to pair with the unique access key.

Last modified: 26 November 2024