AI-powered plotting gallery. One spec, 15 libraries across Python, JavaScript, R and Julia — matplotlib, plotly, ggplot2, D3, Makie & more. Compare and copy working code, or adapt a plot to your own data with an AI agent.
Repo MarkusNeusinger/anyplot, read from GitHub 1 min ago. MarkusNeusinger has not joined Take My Repo.
The agent chat ("use with my data") learns to ask instead of guessing: dates it cannot read, sum or mean, and an order of day and month it only guessed. A selection or a top N becomes a view of the plot, the assistant's figures are checked before a reply goes out, and a spec page names the works it builds on. The about page says who built the site, and a decision log records the owner's calls. 84 PRs (#13056 to #13689), plus 11 Dependabot bumps and 463 pipeline PRs; the full record is CHANGELOG.md, and each PR carries its own reasoning.
.show_all() brings the rest back. draws 10 of 248 rows · Jahr = 2024 · top 10 by Umsatz (#13112).docs/decisions/ holds one file per decision of the owner, with what was proposed, what he decided and its source (#13186)./debug, for 180 days (#13181).stayed_10s) (#13100)./agent without a plot explains it (#13163)./update command. A locally written implementation goes through the pipeline's adopt mode instead (#13129).agent_open and agent_plot_rendered stay (#13060).38.790* had come out as 3879 (#13056).Full Changelog: v4.1.0...v4.2.0
The fine-tuning after the public launch. The agent chat ("use with my data") gets one message box for data and text, a data step that derives columns, and a thread that says what each turn changed. The stats page counts machine readers and shared links, the review pipeline starts recording what it used to assume, and all 323 MUI X implementations now run on MUI X 9. 98 PRs (#12546 to #13058), plus 17 Dependabot bumps and 406 pipeline PRs; the full record is CHANGELOG.md, and each PR carries its own reasoning.
compute, diff, rolling and blank; in the persona tests, five of ten personas had got no plot because a column could not be derived (#12597).split, extract, strip and replace, with literal texts only, never a pattern (#12556)..example_data() in the agent chat. One click asks the assistant for a made-up table to try the plot (#12553).?, by the server's word (#12826).kept: y = 7 line · x label · new: log scale on y (#12826).drawing your plot · 0:42, the try, and the place in the queue (#12950)..use_my_data() → on the spec hub and in the code tab (#12549)..confirm() is gone; .create_plot() confirms the data and starts the plot (#12557)./agent for everyone; the admin's chat looks and behaves like a visitor's (#12589).AI_AGENTS names 18 more crawler tokens and a new seo purpose (#12868, #12962).bots.anyplot.ai. It saw about 1.7 % of real crawler reads; the machine-read lines replace it (#12962).og_image_view no longer goes to Plausible. It carried the fetcher's address and user agent there and measured little (#13055).GET /agent/session/bundle. No page has called it since the public chat dropped .save() (#12555).a:7,b:12,c:5, are data now (#13019).The agent chat ("use with my data") is open to every visitor, as an experimental feature: paste your data on a plot page, and an AI adapts that plot's code to it. 39 PRs (#12252 to #12543); the full record is CHANGELOG.md, and each PR carries its own reasoning.
/agent, experimental. The chat for every visitor, with the experimental notice and its entry points (#12528)./agent, shipped dark. Anonymous tab ids salted per process, a separate kill switch, admin-only routes kept out (#12527)./debug. Runs, tokens and limits per day for the operator (#12518)..use_my_data() →, mirroring .compare() (#12539).roles/editor (#12525).Full Changelog: v3.6.0...v4.0.0
The window since v3.5.0 holds 253 merged pull requests (#12160 to #12506). The release takes the "Use with my data" agent chat, which only admins can use for now, to all 15 catalogue libraries, with the fixes from the user test of 2026-11-01. The full record is CHANGELOG.md, and each PR carries its own reasoning.
plot.tsx mounts as a React component; a MUI X 7 file is served coupled, never blocked (#12483).plot.html. A self-contained page whose plot code runs in an opaque origin (#12480).plot.jl renders in exec mode under the preloaded exec filter (#12464).plot.R renders through the renderer's R row in exec mode (#12448)..plot_all() in the admin agent chat: one table, a plot in every library at once. A grid of cards, through the normal run queue (#12417).CLAUDE_AUTH. All 13 Claude steps in the 8 workflows (#12487).REGEN_PAIRS_PER_DAY. The N oldest (spec, library) pairs across the catalogue (#12487).source_ref on impl-generate.yml and bulk-generate.yml. Code written outside the pipeline reaches main only through the Opus review (#12274).# Data block that ends in a tidy df, one column constant per spec role (#12261).core/adaptability.py; the pipeline does not run it yet (#12244).scripts/site-smoke.mjs, after a vite bump left every page blank for about 45 minutes (#12236).action_required (#12237)./debug/agent asks first (#12502).wait-turn step in all four Cloud Build configs (#12222)."$1,200" reads as 1200, with a warning (#12228).impl-generate.yml fails such a generation, impl-merge.yml refuses such a PR (#12240).execve and execveat (#12410).ctypes.cast passes the plot guard (#12247).Run E2E Tests logs in to GCP (#12429).feature/**. Feature branches get CI through their pull requests (#12441).Full Changelog: v3.5.0...v3.6.0
The window since v3.4.0 holds 57 merged pull requests (#12148 to #12209). Together with the 42 agent network PRs that v3.4.0 left for this release (#12099 to #12146), the release covers 99 PRs. The release is the "Use with my data" agent chat, which only admins can use for now. The full record is CHANGELOG.md, and each PR carries its own reasoning.
/debug. Sessions, plots shipped, kept and repair rates over 7 or 30 days (#12205).agent_feedback table (#12142)./debug/agent in front of the agent chat BFF (#12114)./v1 flow and exits non-zero when the pass rate drops (#12116).PlotResult (#12111).aiplatform.googleapis.com once the month's cost reaches the budget amount (#12109).-basic spec (#12185).anyplot-renderer and anyplot-agents run in Cloud Run (europe-west4) (#12127).agents-eval extra. Removes 29 packages from uv.lock and two Dependabot alerts (#12146).Full Changelog: v3.4.0...v3.5.0
The window since v3.3.0 holds 284 merged pull requests (#11839 to #12147). This release covers the review and scoring work and the general fixes; the agent network from the same window follows in v3.5.0. The full record is CHANGELOG.md, and each PR carries its own reasoning.
line-tanabe-sugano. The first spec reviewed entirely on Opus (#12038, #12043).daily-regen regenerations becomes a dated report (#12055)..env is read only from the checkout root. (#12105)Full Changelog: v3.3.0...v3.4.0