Generating matplotlib charts from the terminal

Turn numbers into line, bar, pie, scatter, histogram, and pareto PNG charts using natural language or Arka CLI commands, powered by matplotlib.

Turn numbers into PNG charts. Seven chart kinds supported via offline NL parsing — no LLM required for most requests.

Always use arka in front. Every example below is an arka … command or arka "natural language". In Fish you can also use bundled shortcuts like chart …, but docs and copy-paste examples use arka so routing is explicit.

pip install "arka-agent[charts]"

Chart types

Kind	CLI	Natural language (via arka)
line	arka chart line AAPL TSLA --range 1y	arka "compare Apple and Tesla stock last year"
bar	arka chart bar --data "A:1,B:2"	arka "phone sales Apple 230 Samsung 210"
pie	arka chart pie --data "A:40,B:30"	arka "pie chart traffic organic 400 direct 300"
scatter	arka chart scatter --data "100:200,120:190"	arka "scatter ad spend vs revenue 100 200 120 190"
histogram	arka chart histogram --data "12,15,18,22"	arka "histogram response times 12 15 18 22"
pareto	arka chart pareto --data "A:45,B:28"	arka "pareto defects Scratches 45 Dents 28"
prediction	arka chart predict AAPL --days 30	arka "predict AAPL stock price chart next 30 days"
model traits	arka chart models --type scatter	arka "chart AI model characteristics scatter latency vs score"
table image	arka chart table sales.csv --title "Q1"	arka "render metrics.csv as table png"
from file	arka chart from metrics.csv --type line	arka "chart ~/data/metrics.csv as line"

Dataset columns (flexible NL)

Any column names work — Arka parses axes from phrasing, not fixed keywords:

arka "present time vs income from income.csv"
arka "plot monthly_revenue over fiscal_period from sales.csv"
arka "visualize payroll.csv with tenure and salary"
arka "chart from sales.csv by region value revenue"

Supported patterns: Y over X, X vs Y, by X value Y, with X and Y, x-axis X y-axis Y. Time-like columns auto-route to line charts; other vs pairs default to scatter.

Natural language examples

Pass the phrase as a single quoted argument to arka:

arka "make a chart of training loss"
arka "chart training loss 1:0.8 2:0.5 3:0.3"
arka "plot loss over epoch"
arka "plot X over Y 1:10 2:20 3:30"
arka "visualize metrics.csv"
arka "chart from csv sales.tsv as bar"
arka "graph my data"

Direct CLI

arka chart line AAPL TSLA --range 1y
arka chart predict RELIANCE.NS --days 30 --range 1y
arka chart bar --data "Apple:230,Samsung:210,Xiaomi:140" --title "Phone sales"
arka chart pie --data "Organic:400,Direct:300,Referral:300" --title "Traffic"
arka chart table sales.csv --title "Q1 sales" --max-rows 25

Table as PNG

Render a spreadsheet-style table (headers + rows) as a PNG — useful for slides, docs, or sharing CSV previews:

arka chart table sales.csv
arka chart table metrics.csv --title "Experiment results" --max-rows 30 -o ~/Desktop/table.png
arka chart from report.csv --type table

Natural language:

arka "render sales.csv as table png"
arka "table image from metrics.csv title 'Results'"

Chart defaults

Store reusable data in ~/.config/arka/charts.yaml:

defaults:
  scatter:
    data: "100:200,120:190,170:280"
    xlabel: "Ad Spend"
    ylabel: "Revenue"
  line:
    tickers: [AAPL, TSLA]
    range: 1y

Charts save to ~/Pictures/arka-generated/ and open automatically on macOS/Linux.

Future prediction (all domains)

Beyond stock tickers, Arka predicts futures from live data sources:

Domain	Example	Data source
Weather	arka predict future "weather in Delhi next 7 days" --chart	Open-Meteo
Rainfall / monsoon	arka predict future "monsoon rainfall in Mumbai" --chart	Open-Meteo + ISRO MOSDAC notes
Stock	arka chart predict AAPL --days 30	Yahoo Finance
Satellite	arka predict future "ISS pass over Bangalore"	Open Notify
Kaggle CSV	arka predict future "from kaggle owner/dataset next 30 days" --chart	Kaggle API + trend
Local CSV	arka predict future "forecast from metrics.csv" --chart	Local file trend

ISRO GSMaP satellite rainfall archives: set MOSDAC_USER and MOSDAC_PASSWORD in .env (see MOSDAC API manual). Charts work without MOSDAC; credentials unlock HDF5 archive search.

arka predict future "rainfall in Mumbai next 2 weeks" --chart
arka predict future "from kaggle heptapod/titanic next 30 days" --chart
arka chart predict --domain rainfall "monsoon in Pune" --days 14

Natural language:

arka "predict monsoon rainfall in Mumbai next 2 weeks"
arka "forecast kaggle dataset heptapod/titanic next 30 days with chart"

Kaggle datasets need API credentials (KAGGLE_USERNAME / KAGGLE_KEY or ~/.kaggle/kaggle.json). See Kaggle datasets.

AI model characteristic charts

Compare LLM models with scatter plots (latency vs score) or bar charts (scores / local model size):

arka benchmark run --dry-run          # populate benchmark-results.json first
arka chart models --type scatter    # latency vs score with labeled points
arka chart models --type score_bar --profile chat
arka chart models --type size_bar   # local Ollama models by parameter scale

Natural language:

arka "chart AI model characteristics with data points"
arka "scatter plot of model latency vs benchmark score"
arka "graph LLM model comparison"

# CHART_OUTPUT_DIR=~/Pictures/arka-generated
# CHARTS_CONFIG=              # override defaults file path