Telemetry & Results

Telemetry, results tracking, results.tsv.

Overview

The results module (results.py) tracks experiment outcomes in results.tsv. The telemetry module (telemetry.py) provides run-level metrics and cost tracking.

Results Tracking

Each experiment appends a row to results.tsv with these columns:

Column	Description
commit	Git commit hash (if kept) or -- (if discarded)
metric	Measured metric value
guard	Guard result or -- if no guard configured
status	kept or discarded
confidence	Statistical confidence score
description	One-line summary of the change

Result Queries

• read_results() — Load all results for a marker
• get_latest_metric() — Current best metric value
• get_kept_metrics() — All successful improvements
• append_result() — Record a new experiment outcome

Telemetry

The telemetry.py module parses Claude Code’s --output-format stream-json into a TelemetryReport dataclass covering tokens, cost, tools used, and errors.