EXTRACT01 / 07
EXTRACT
Pulls clean data out of any document.
When data is locked in an image or report.
Agents that turn months of chart work into days.
ways to use
Either you tell them what you want.
Or you let them tell you.
ask mode
Ask for a chart. Get a good one.
›
explore mode
Add your data. Get the insights.
arr-events.json
account_id
event
arr_delta
inside
What's actually doing the work.
Seven agents, each owning one move from raw data to finished chart. Click any to see what it does.
benchmarks
Why the agents beat a raw model call.
Same model everyone else calls. The harness around it is the difference, and it's measurable.
We benchmarked Graphy's mutate-only agent against direct LLM calls (GPT-5.5, Claude Opus 4.8, Gemini 3.5 Flash) on 24 fixed table tasks: 6 operations (sum, count, mean, max, pivot, unpivot) across 4 file sizes (~100 to ~30k rows), each on 3 seeds. Outputs are checked in code against precomputed ground truth (columns, row count, values within tolerance). Pass means 100% correct. No partial credit, no human judges.
Numbers you can actually trust.
Every request runs through a fixed harness so your users never see a chart the model invented.
build with it
A handful of imports. Your call.
Call one, chain a few, or hand one to an agent.
graphy-demo.ts
npm i @graphysdk/agents-sdk
1
2
3
4
5
6
7
8
9
10
11
12
13
// one call — a chart your user can read.
import { GraphyAiSdk } from '@graphysdk/agents-sdk'
import { GraphProvider, Graph } from '@graphysdk/core'
const ai = new GraphyAiSdk({ apiKey, baseUrl })
const { config } = await ai.generateGraph({
config: { type: 'column', data: stripeExport },
userPrompt: 'Plot revenue by quarter, mark the Q1 spike',
})
// → validated + themed, ready to ship.
<GraphProvider config={config} theme={brand}><Graph /></GraphProvider>
shippedreturns GraphConfig · 1.2s