Terminal Line Plot

Terminal Line Plot

Use this skill to turn numeric sequences into terminal-readable line charts. It is task-agnostic: do not assume any domain, metric direction, filtering rule, or data layout unless the user or artifacts define it.

Workflow

1. Identify the series to plot:
   • x values are usually ordered steps, dates, versions, generations, epochs, or run ids.
   • y values must be numeric.
   • If multiple plausible metrics exist, choose the one the user asked for; otherwise state the selected metric before plotting.
2. Preserve metric semantics:
   • Use exact metric names from artifacts when available.
   • Label whether higher or lower is better if known.
   • Do not mix partial and complete points silently; mark provisional or incomplete points in the title, notes, or point list.
3. Use scripts/plot_series.py for deterministic plotting.
4. Print a compact table or point list when it prevents misreading the curve.
5. Do not write image files unless the user explicitly asks. Prefer stdout.

Script Usage

Run the bundled script from the skill directory:

python /path/to/terminal-line-plot/scripts/plot_series.py \
  --points "0:-4.49,1:-4.53,2:-4.87,3:-4.77,4:-3.23" \
  --title "Best future_fitness by generation" \
  --x-label generation \
  --y-label future_fitness \
  --higher-better

Input options:

• --points "x:y,x:y,..." for quick point lists.
• --csv FILE --x-field FIELD --y-field FIELD for CSV.
• --json FILE --x-field FIELD --y-field FIELD for JSON arrays of objects, arrays of pairs, or mappings.
• --stdin csv|json|points to read data from stdin.

Useful options:

• --height N controls chart height.
• --x-step N controls horizontal spacing between points.
• --provisional-x X may be repeated to mark incomplete points with ○.
• --value-table prints the plotted values below the chart.
• --invert-y-note only changes the note; it does not invert the axis.

Plotting Guidelines

• Prefer line drawing characters ●, ○, ╱, ╲, and ─.
• Keep the x-axis readable. If labels collide, increase --x-step or plot fewer points.
• Use a taller chart for noisy or close-valued curves.
• For negative scores where "higher is better", the visually higher point should still mean numerically larger.
• If the active/final point is incomplete, mark it provisional and do not draw conclusions from it.
• For dense data, aggregate intentionally: best per generation, mean per epoch bucket, rolling median, or another stated rule.

Example Patterns

Manual points:

python /path/to/terminal-line-plot/scripts/plot_series.py \
  --points "0:7.88,1:7.78,2:8.45,3:8.61,4:9.28" \
  --title "Mean objective value by generation" \
  --y-label "objective" \
  --value-table

CSV:

python /path/to/terminal-line-plot/scripts/plot_series.py \
  --csv metrics.csv \
  --x-field generation \
  --y-field score \
  --title "Score by generation"

JSON:

python /path/to/terminal-line-plot/scripts/plot_series.py \
  --json metrics.json \
  --x-field gen \
  --y-field best_score