

The TaskForge Analytics & Insights module gives you full visibility into how your pipelines, models, and tasks are performing over time. From high-level operational dashboards to granular per-request telemetry, you have the data you need to optimize performance, debug issues, and make informed infrastructure decisions.
The main analytics dashboard is accessible from the left sidebar under Analytics. It provides a real-time overview of your workspace activity, including:
All metrics can be filtered by date range, pipeline, model, or team. The dashboard updates every 30 seconds in real time.
For each pipeline, TaskForge tracks a detailed set of execution metrics. To view pipeline-level analytics:
Key metrics available per pipeline include execution duration, step-level breakdown, retry counts, error frequency by step, and data throughput in bytes per second. You can export any metric view as a CSV or PNG chart for reporting purposes.
For AI-powered pipelines, the Models analytics panel provides inference-specific metrics including:
Accuracy drift alerts can be configured to notify your team via Slack or email when output confidence drops below a defined threshold.
TaskForge allows you to build custom reports by combining any available metric with filters, groupings, and visualization types. Reports can be:
Custom reports support bar charts, line graphs, scatter plots, and summary tables.
The Log Explorer provides full-text search across all task execution logs within your workspace. You can filter logs by pipeline, task ID, status code, time range, and custom log levels.
Structured logs emitted by your tasks using <taskforge.log()> are automatically indexed and searchable within seconds of execution. This makes the Log Explorer the fastest way to debug a failing task or trace a specific data record through your pipeline.
TaskForge includes a built-in alerting engine that monitors your pipelines for anomalies and notifies you when thresholds are breached. To configure an alert:
Alerts support both static thresholds and dynamic anomaly detection, which learns from historical patterns to flag unusual behavior automatically.