Data Visualization

D3113.9k★

D3 gives you complete control over custom data visualizations, building charts and interactive graphics from scratch that no charting library offers out of the box. It's not a charting library. It's a visualization toolkit that binds data to DOM elements and lets you manipulate them directly, the most-starred data visualization library on GitHub. ISC license (permissive). Created by Mike Bostock, now maintained by Observable. D3 powers visualizations at the New York Times, NASA, Uber, and thousands of other organizations. The power is in the primitives: scales that map data to pixels, axes, shape generators, geographic projections, force simulations, hierarchy layouts. You compose these to build exactly what you need. The Observable notebook ecosystem has thousands of examples you can fork. Fully free. No paid tier for D3 itself. Observable (the company) offers a paid collaboration platform, but D3 the library is and always will be free. Solo developers: free, but expect a steep learning curve. D3 is not drag-and-drop. Small to large teams: if you have a developer who knows D3, the results are unmatched. If nobody on the team wants to learn it, use Chart.js or ECharts. The catch: D3 is powerful precisely because it's low-level, and that's also why it's hard. Making a simple bar chart takes 50 lines of code instead of a config object. The learning curve is real, weeks, not hours. For standard charts, use a higher-level library. Reach for D3 when nothing else can do what you need.

worldmonitor90.2k★

World Monitor is a live world-events dashboard: conflicts, disasters, flights, ships, markets and news on a 3D globe or flat map, with AI-written briefs on top. The code is AGPL-3.0 and free to run yourself, and the hosted site has a free tier. Trying it is clone, npm install, npm run dev, with no environment variables. A useful install is more work: the Docker stack wants Node 22, Redis, four generated secrets and a seed-script run, because panels stay blank until the seeders fill Redis. Many feeds need their own free API keys, and the briefs run on Ollama or any OpenAI-compatible endpoint. The hosted Free plan gets core panels and 3 dashboard tabs. Pro is $39.99/month or $359.99/year for the analyst chat, digests and alerts, licensed to one person. Client work needs Pro Business at $49.99/month; the API runs $99.99 to $299.99/month. Solo: free or self-host. Small teams: self-host with your own keys. Larger teams: Enterprise, priced by sales. The catch: the AI briefs are leads to verify, not intelligence. And self-hosting frees the code, not the data. Every upstream source keeps its own terms, and Cloudflare's outage feed is paid.

Chart.js67.8k★

Pass it data and options, get a canvas-rendered chart. Done. This is the most popular charting library on the web for a reason: it works, it's simple, and it looks good out of the box. Animation, tooltips, legends, responsive sizing, all built in. Eight chart types cover 90% of dashboard needs. MIT, fully free. No paid tier. Has been around since 2013 and isn't going anywhere. The catch: Chart.js renders to canvas, not SVG. That means you can't style individual elements with CSS, can't easily make charts accessible to screen readers, and complex interactions (clicking on specific data points, custom hover behaviors) require more work than SVG-based libraries. If you need highly customized, interactive data visualizations, D3.js gives you full control. If you need standard charts that look good fast, Chart.js is the answer.

ECharts67.7k★

ECharts renders everything from simple bar charts to complex 3D globes, geomaps, and real-time streaming dashboards, all client-side with no server dependencies. It handles everything from a simple dashboard widget to a full geographic heatmap with 100K data points. Apache Foundation project, TypeScript core. The feature set is massive: 20+ chart types, animations, responsive layouts, dataset transforms, and a visual builder at echarts.apache.org that lets you configure charts before writing code. It handles real-time streaming data and lazy-loads chart types you don't use. Everything is free. No paid tier, no hosted version, no premium charts behind a paywall. Apache 2.0 license means you can use it in commercial products without restrictions. Solo developers get a production-grade charting library for $0. Teams of any size get the same thing. The community is enormous. StackOverflow answers for almost any question. The catch: the API surface is huge. Configuration objects get deeply nested and the docs, while comprehensive, assume you already know what you want. If you need something quick and simple, Chart.js gets you there faster. If you need advanced interactivity or large datasets, ECharts is the better pick.

Streamlit46.2k★

Streamlit turns Python scripts into web apps in minutes: dashboards, demos, internal tools. No HTML, no CSS, no JavaScript. You write Python, and Streamlit gives you a web interface with charts, tables, sliders, and file uploads. Apache 2.0, Python, owned by Snowflake. The workflow is dead simple: write a Python script using Streamlit's API (st.write, st.chart, st.slider), run streamlit run app.py, and you have a web app. It re-runs the script top-to-bottom on every interaction, which is brilliant for prototyping and awkward for complex state management. The library is fully free. Streamlit Community Cloud hosts your apps for free with generous limits: public apps, GitHub-connected deployment. Snowflake offers Streamlit in Snowflake for enterprise data apps. Solo data scientists: this is your tool. Build demos, share analyses, prototype ML interfaces. Small teams: great for internal dashboards. Medium to large: works for internal tools, but production customer-facing apps will outgrow the re-run-on-every-click model. The catch: Streamlit re-executes your entire script on every user interaction. For simple apps, fine. For apps with expensive computations, you need caching decorators everywhere. The layout system is limited: you get a single column by default and basic column/tab options. If you need a real web app with complex navigation and state, you've outgrown Streamlit.

kibana21.4k★

Kibana is the visualization layer for Elasticsearch. If your data lives in an Elastic cluster, this is how you build dashboards, run queries, and monitor everything from logs to APM traces. The UI handles everything from ad-hoc searches to complex aggregations without writing code. Self-hosting is free under the Elastic License 2.0, but running Elasticsearch underneath is the real commitment. Plan for a dedicated cluster, index management, and ongoing maintenance. This is not a weekend project for small teams. Solo developers and small teams doing log analysis can get by with the free tier and a managed Elastic Cloud instance (starts at $95/mo). Enterprise teams running their own clusters get the full power but need dedicated ops. The paid tiers add security features (SSO, field-level permissions), machine learning, and cross-cluster search. The catch: Kibana without Elasticsearch is nothing, and Elasticsearch at scale is expensive to operate. If you just need dashboards over Postgres or CSV data, look at Metabase or Grafana instead.

Mapbox GL JS12.5k★

Mapbox GL JS renders interactive, GPU-accelerated maps in the browser that you can tilt, rotate, fly through, and layer custom data on. It renders vector tiles using WebGL, which means smooth 60fps panning and zooming with 3D building extrusions, custom styling, and data visualization layers. Here's where it gets complicated. Mapbox GL JS v1 was open source (BSD). Version 2+ switched to a proprietary license. You can view the source code, but you can't use it without a Mapbox access token, and that token ties you to Mapbox's pricing. Free tier: 50,000 map loads/month. After that, $5 per 1,000 loads. There's no self-hosting option for v2+. You need Mapbox's tile servers and API. The library itself loads from npm but phones home for tiles and authentication. Solo developers and small projects: the free tier is generous. 50K loads covers most side projects and small apps. Medium teams: costs become real. A site with 500K map loads/month runs ~$2,250/mo. That adds up fast. The catch: the license switch burned the community hard. MapLibre GL JS forked from the last open source version and is actively maintained. If you want truly free maps, start there instead.

lightdash6.2k★

Lightdash is an open source BI tool that turns your dbt models into self-service dashboards. You define metrics in YAML right next to your data, and your team explores them and builds charts without writing SQL. The positioning is blunt and accurate: an open source Looker. The core is MIT-licensed and free to self-host with full BI functionality. The deployment is well-documented, Docker, Kubernetes via Helm, or a one-click Render setup, but you are still running the app, its database, and the connection to your warehouse. Call it moderate ops: not heavy, not nothing. The bigger dependency is dbt. Lightdash is built around it, so if your team does not model data in dbt, most of the value evaporates before you start. For data teams already invested in dbt who want governed, self-service analytics without Looker's price tag, this is a strong fit. Solo and small teams: self-host the MIT core, free. Larger orgs that want managed hosting or the enterprise features: Lightdash Cloud is the paid path, and the enterprise code in the /ee directory is not open source. Not a dbt shop? Look at Metabase, which connects straight to databases without a modeling layer. The catch is that dbt coupling. It is the reason Lightdash is clean and governed, and also the reason it is useless to teams that model data some other way. Check that box first; everything else follows from it.

pygraphistry2.6k★

PyGraphistry is the Python client for Graphistry, a GPU-accelerated graph visualization platform. Load a few million nodes and edges, get an interactive zoomable graph in your browser. Free Python library, paid server tiers. The library is open source and easy to use. Pandas DataFrame in, visualization out. You point it at the Graphistry Hub (free tier with limits) or your own enterprise server. The rendering happens server-side on a GPU, which is what makes this scale to graphs that would melt D3 or sigma.js. For data scientists doing exploratory analysis on biggish graphs, this is the right tool. The free Hub tier is fine for personal use and demos. Teams that need private data or higher limits pay for the enterprise server, which is real money. Quote-required, not self-serve. If you can't justify that, look at Memgraph Lab, Neo4j Bloom on Aura's free tier, or roll your own with sigma.js. The catch is the GPU dependency. The client library is free but useless without a Graphistry server, and you can't self-host the server free tier. Plan around that before you build it into a product.

RustGIS2.4k★

RustGIS is a desktop GIS (mapping and spatial analysis software, what planners and ecologists use) written in pure Rust. It runs natively on Windows, macOS and Linux, or in a browser tab. It's free under Apache 2.0, and it comes from Qiusheng Wu, the University of Tennessee professor behind leafmap and geemap. Real pedigree. Nothing to host. Grab an installer (or winget / brew) and open a Shapefile, GeoPackage, GeoParquet or cloud-optimized GeoTIFF. ArcGIS Pro users will recognize the ribbon. Every button is a command you can script from rustgis-cli or Python, or hand to an AI assistant using your own model key. Over 1,000 terrain, hydrology and LiDAR tools come from the GeoLibre and Whitebox suites. Solo, small teams, larger teams: all free. Worth an afternoon if you pay for ArcGIS Pro just to make maps. For production work, QGIS is still the free default. The catch: the repo is five days old and on version 0.2. Its own roadmap lists vertex editing, spatial join, intersect and union as not done yet, and those are everyday GIS operations. The feature list reads like ArcGIS Pro; the bug tracker reads like a first release. Try it, don't migrate.

vizzu-lib2.0k★

Vizzu does something no other charting library does well. It's built around the idea that data stories are told through transitions, not static images. You feed it a dataset and describe how you want it visualized. Change the description, and it animates the transition. The same data can flow from a grouped bar chart to a treemap to a bubble chart with physics-based animation. It's impressive to watch and makes presentations and dashboards feel alive. Completely free under Apache 2.0. No paid tier, no premium chart types, no watermarks. For standard charting, Apache ECharts or Chart.js are the workhorses. For custom data art, D3 is still king. Vizzu fills the niche of animated data storytelling that nothing else does as cleanly. The catch: this is a presentation and storytelling tool, not a drop-in replacement for Chart.js or D3. If you need standard dashboards with tooltips, drill-downs, and real-time updates, Vizzu isn't built for that. The API is also different from what you're used to. It's not "make me a bar chart," it's "here's data, here's how to arrange it." There's a learning curve. The community is small and documentation could be better.

meta2d.js1.3k★

meta2d.js is a 2D canvas engine for live data screens: the animated dashboards that show a factory floor, a power grid, or an IoT fleet updating in real time. If you have seen a control room where valves and meters move as the numbers change, this builds that in a browser. The engine is MIT and free. You embed it in your own web app, bind it to your live data source, and it handles the rendering, animation, and interaction. It is a library, not an app, so the setup is however much scaffolding your project already has. The docs and examples are solid but heavily Chinese-first, which slows you down if you do not read it. The catch on money: the engine is free, but le5le sells the visual editor and the pre-built component libraries that make it fast to actually design a screen. Solo builders and small teams can hand-code against the free engine. If you are shipping SCADA dashboards for clients, the paid editor pays for itself in saved time. This is a niche tool. If you just need charts, use a charting library. meta2d earns its place only when you need interactive, animated, real-time diagrams that a chart cannot express: digital twins, monitoring floor plans, industrial control views.

shaper1.3k★

Shaper turns SQL queries into dashboards. Write a query, tag columns with chart types like ::BARCHART_STACKED, and it draws the chart. It runs on DuckDB, so it can query Postgres, MySQL, S3 files, CSV, Parquet and Excel. Embedded analytics with row-level security, white-labeling, PDF and Excel exports and scheduled reports are all in the free MPL-2.0 version. docker run --rm -it -p5454:5454 taleshape/shaper gets you running. Production means owning TLS, backups, upgrades and patching yourself. Dashboards live in Git, and agent skills let Claude Code or Cursor write them. Taleshape sells managed hosting, still labeled beta: Standard $59/month, Pro $249/month (adds OIDC and proxy-header SSO, an MCP server and usage analytics), Enterprise $799/month (audit logging, HA, data engineering help). Yearly is just 12x monthly. Solo and small SQL teams: free. Teams that need Okta-style login: Pro, or build your own JWT login endpoint for free. The catch: the UI has no view-only roles, so viewers come in through embeds or share links. That share path was the hole in an October advisory: a public-share token could deploy dashboards and run SQL on the server, fixed in 0.25.2. Patch first. Analysts who don't write SQL should look at Metabase.