sf-ai-agentforce-observability

You can extract Agentforce session tracing data from Salesforce Data Cloud by running the stdm-extract CLI, which automates pulling sessions, interactions, steps, and messages into Parquet files for efficient storage and analysis.

Polars analysis processes Agentforce STDM data by querying extracted Parquet files to detect performance patterns, bottlenecks, and escalation paths across agents using the STDMAnalyzer Python API.

Yes, this supports high-volume Agentforce tracing workloads by enforcing STDM schemas and enabling both incremental and full extractions to Parquet, ensuring scalable data processing across large session datasets.

The best way to debug Agentforce sessions is to extract STDM tracing data to Parquet using the CLI, then analyze session interactions and steps with the Polars-driven STDMAnalyzer to identify execution bottlenecks.

Yes, STDM extraction enforces schema validation for Agentforce data by applying STDM schemas across sessions, interactions, steps, and messages to ensure consistent data structure during Parquet storage and analysis.

Parquet-based storage for Agentforce trace analysis requires Python dependencies like pyarrow and polars, and while it enables efficient columnar analysis, it does not support real-time streaming of session tracing data from Data Cloud.