Pipelines & Runs

A pipeline turns a source into intelligence. Each pipeline ingests conversations, analyzes them, applies your rules, and materializes results that downstream views and destinations read.

Stages of a run

1. Ingest — pull new conversations from the source (idempotent, so re-running only processes what is new).
2. Analyze — transcribe if needed, then extract signals, sentiment, topics, stakeholders, and next steps.
3. Apply policy — run your rules and overrides to produce final results.
4. Materialize — write results to account memory and the views that read them.

Schedule and trigger

Run a pipeline on a schedule or trigger it on demand from Settings → Pipelines. Because data is stored at every layer, you can re-run from any layer without redoing everything.

Monitor

Every run is recorded with its inputs, outputs, and any rule firings, so you can see exactly what happened and re-run if needed.