AI Observability: Jobs

The AI Observability content module includes 18 scheduled jobs. One maintains the GPU asset in Inventory, eight populate Claude Code lookups, three register subscription-limit incidents, and another six are disabled examples of using the api command to query equivalent-report lookups.

Job	Purpose
ai_observability_collect_gpu_asset	populates the gen_ai_gpu_asset index from DCGM Exporter metrics for the GPU asset
ai_observability_lookup_analytics_daily	populates claude_code_analytics_daily - an equivalent of the Analytics API daily member summary
ai_observability_lookup_analytics_models	populates claude_code_analytics_models_daily - tokens and costs by model
ai_observability_lookup_analytics_tools	populates claude_code_analytics_tools_daily - accepted and rejected edits by tool
ai_observability_lookup_sessions	populates claude_code_sessions with sessions from the file-based data source
ai_observability_lookup_sessions_otel	adds sessions for members using telemetry only to claude_code_sessions
ai_observability_lookup_activity	populates claude_code_activity_daily - an equivalent of the Compliance API audit feed
ai_observability_lookup_spend_report	populates claude_code_spend_report - an equivalent of the subscription spend report for completed days
ai_observability_lookup_identity	builds claude_code_identity - a mapping of OS account and host to member address
ai_observability_alert_limit_weekly_80	registers an incident when a member has used at least 80% of the weekly subscription window limit
ai_observability_alert_limit_exhausted	registers an incident when a member has used the full subscription window limit
ai_observability_alert_limit_collector_silent	registers an incident when a member's limit snapshots have not arrived for more than a day
ai_observability_api_activity	sample query to claude_code_activity_daily - an equivalent of the Compliance API Activity Feed
ai_observability_api_analytics_daily	sample query to claude_code_analytics_daily - an equivalent of the Analytics API usage_report/claude_code
ai_observability_api_analytics_models	sample query to claude_code_analytics_models_daily - an equivalent of the model_breakdown section
ai_observability_api_analytics_tools	sample query to claude_code_analytics_tools_daily - an equivalent of the tool_actions section
ai_observability_api_sessions	sample query to claude_code_sessions - an equivalent of the Compliance API session list
ai_observability_api_spend_report	sample query to claude_code_spend_report - an equivalent of the subscription spend-report export

Examples of incidents registered by the limit-monitoring jobs are on the Incident Manager integration page.

Populating Lookups

The equivalent-report lookups follow the structure of Anthropic reports but are populated by module jobs from collected telemetry. On each run, the jobs recalculate data for the last two days, the spend-report job for the last eight completed UTC days, and the claude_code_identity lookup job for the last 30 days. Earlier telemetry is not added to the lookups: analytics accumulate from the first runs after installation.

The jobs do not populate claude_code_members or claude_code_seat_pricing; an administrator maintains these lookups. The lookups are listed in Data Sources.

Jobs Using the api Command

The six ai_observability_api_* jobs are disabled and do not change data. They show how to retrieve rows from an equivalent-report lookup through the SA Data Storage REST API using the api command. Before enabling them:

1. Replace the HOST:PORT placeholder in the request url with the A Data Storage address
2. Add the sme.sa.api.header.sm_self_auth header token to the keystore with the value Basic <base64 of username:password> for a Search Anywhere Framework user, as described in Creating a token

While the url contains the placeholder, the job request fails with main_url in connection_config is null.