poli0981andClaude Opus 5.5 f91ad37e5b fix(pipeline): ASCII appids and CJK-aware description cuts
- Python's digit class also takes full-width and other scripts' digits.
  A queued link in them gave an appid that missed every comparison and
  made rotation.slice_of() raise, stopping every refresh. extract_appid
  refuses a run of digits with a non-ASCII one in it; normalize_link()
  reads NFKC first, so a Japanese keyboard's full-width appid is ASCII.
  The blocklist uses ASCII digits and fullmatch, like removal.ts.
- truncate_description() backs off to just after a CJK sentence end as it
  does to a space, and trims the CJK comma and full stop before the
  ellipsis. Stored descriptions (all within the cap) never change, and
  _fuller_cut() lets Refresh Store Data write a re-cut without calling it
  news.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-30 17:28:28 +07:00
2026-10-30 07:13:30 +00:00
2026-10-30 03:44:24 +00:00
2025-12-19 18:30:23 +07:00

Games Count Last Updated Top Online Version Web App Desktop Android Health Check Visitors Sponsor

A curated list of free-to-play games on Steam — now ~3,600 of them — no quality guarantees, no endorsements, just pure listings by a broke, unemployed Vietnamese dev with too much free time and a Steam library full of sale games (80% under $10).

v4.0.0 is the version that got rebuilt from the ground up. It is still a list. It is also:

  • a web app at https://free-steam-games.win/ (SvelteKit, served by a Cloudflare Worker) with a virtualised table of every game, charts, fuzzy search, filters you can link to, CSV/JSON export, a command palette, a page per game that search engines can read, en/vi i18n, and an installable offline mode. It is read-only for visitors: there is no sign-in;
  • a desktop app (Tauri 2, resizable window, updates itself from GitHub Releases — Win/Mac/Linux installers);
  • an Android app (sideloaded APK, Android 11+);
  • an admin at /admin, behind Cloudflare Access: a review queue for the games the daily discovery sweep finds (approved by the maintainer, or in capped daily batches by a rule the maintainer turns on), corrections to published games, and removals;
  • the same Python pipeline that has run daily under GitHub Actions since v1.

Powered ~70 % by AI (Grok for v1, Claude for v2 + the entire web/desktop layer) — code is reviewed and tested, but mistakes happen. See the Disclaimer for the specific accuracy caveats (genre best-effort, English-only assumptions, test-data leakage, unsigned-commit dev artefacts).

Wallet safe (for now). Play at your own risk.

Surface URL
🌐 Web app https://free-steam-games.win/
🖥️ Desktop releases GitHub Releases (Win .msi, Mac .dmg, Linux .AppImage)
📱 Android APK GitHub Releases (sideload .apk, Android 11+ — see docs/android-support.md)
🏆 Top online (md) games/top-online.md
🎯 Top offline (md) games/top-offline.md
📦 Raw data data/data_001.jsonl … data_005.jsonl (5 sharded JSONL files)
🛠️ About https://free-steam-games.win/about (in-app)
👤 Authors AUTHORS.md — maintainer + contact channels + AI-assistant disclosure

What's new in v4.0.0

Web app at free-steam-games.win/

A SvelteKit 2 / Svelte 5 app, prerendered and served by a Cloudflare Worker that also proxies the dataset (/api/data/*) and Steam artwork (/img/*), so your browser talks to this site only — apart from Cloudflare's analytics beacon and its Turnstile human check, both loaded only after you accept the terms. Features:

  • Browse — virtualised table (TanStack Virtual), filters for genre, type, platform, status, anti-cheat and safety, Fuse.js fuzzy search, sortable columns, 50/100/200/500/all pages. Filters live in the URL, so a filtered view can be shared.
  • Charts — Apache ECharts: dashboard KPIs, genres, platforms, languages, tags, anti-cheat (chart and a searchable index), reviews, player tiers, release years and catalogue growth, DRM/DLC, delisted games, plus a stats page (Metacritic, retention, games that went quiet).
  • A page per game — /games/:appid is prerendered for search engines and filled with live player counts and reviews in the browser.
  • Fresh data, verified — data/index.json carries a SHA-256 per shard; the app only caches a generation it could verify, keeps the last good one offline, and re-checks when you come back to the tab.
  • Installable — a web app you can install, with an offline mode and an update prompt.
  • i18n vi / en — auto-detected, switchable from the welcome page or Settings.
  • ⌘K command palette, first-run introduction, and the legal documents rendered in the app.

Pushing to main deploys (Cloudflare Workers Builds). Pipeline commits to data/** do not need a rebuild: the data is proxied, not bundled. See docs/DEPLOYMENT.md.

Desktop and Android apps (Tauri 2)

A native shell around the same web app, distributed via GitHub Releases:

  • Windows .msi, macOS .dmg (universal Apple Silicon + Intel), Linux .AppImage and .deb, and an Android .apk.
  • Resizable window; steam:// links open the Steam client.
  • The desktop app checks for a signed update on start and installs it when you say so, keeping your settings.
  • Same data path as the web app (/api/data/* on free-steam-games.win), same offline cache.

Built by .github/workflows/release-desktop.yml (matrix Win/Mac/Linux via tauri-apps/tauri-action) and release-android.yml, triggered by a desktop-v* / android-v* tag or a manual run. Each release is created as a draft.

# from web/, locally:
npm run tauri:dev      # dev build with hot reload
npm run tauri:build    # release binary into web/src-tauri/target/release/bundle
The admin at /admin

Discover New Games sweeps the Steam store daily and proposes candidates into a Cloudflare D1 queue. The maintainer reviews them in a small Svelte app embedded in the Worker (never shipped to visitors): approving commits the store link to scripts/temp_info.jsonl, and the pipeline does the rest, checking every game again (free, a game, reachable) before it is published. Approval can also run on its own, in batches with a daily cap, once the maintainer turns it on. Decided and already-published rows are read-only. Corrections to published games are saved as override files that survive every pipeline run, and a removed game goes into data/blocklist.jsonl, which keeps it out of the catalogue until it is restored. An overview, the pipeline's runs and a data-quality report complete it. See docs/ADMIN.md.

v2.x foundations still here

The Python pipeline that powered v2 is untouched and runs daily as before:

  • Schema v2.2 (23 + manual fields per record), CC BY 4.0 (see LICENSE-DATA), sharded data in data/data_*.jsonl (cap 800 records/shard) plus a data/index.json manifest with a SHA-256 per shard.
  • HTML scraper for accurate language tables, DLC pricing, user tags and the developer's anti-cheat disclosure — single GET per game.
  • Health-checked daily: Steam's store API screens every game, and a game is removed only when appdetails agrees.
  • Reviews refreshed daily from Steam's own summaries; full refresh daily. Every data job queues for its turn and has a time budget, so a growing catalogue slows a refresh down instead of breaking it (plan).
Architecture
.
├── data/                       # Sharded data (the source of truth)
│   ├── data_001.jsonl          # ≤ 800 records
│   ├── data_002.jsonl … 005    # the rest
│   └── index.json              # last_updated, totals, shard manifest
│
├── games/                      # The two generated leaderboards
│   ├── top-online.md           # live leaderboard
│   └── top-offline.md          # most-played offline-capable
│
├── scripts/                    # Python data pipeline
│   ├── core/
│   │   ├── constants.py        # rate limits, anti-cheat patterns, schema
│   │   ├── steam_client.py     # rate-limited HTTP client
│   │   ├── data_store.py       # JSONL I/O, dedup, schema merge
│   │   ├── fetcher.py          # batch processor
│   │   ├── scraper.py          # HTML page parser
│   │   └── health_checker.py   # status detection
│   ├── ingest_new.py / update_data.py / discover_new.py / …
│   └── temp_info.jsonl         # ingest queue
│
├── web/                        # Everything that is not the pipeline
│   ├── src/                    # SvelteKit app: routes, components, charts, i18n
│   ├── admin/                  # /admin review app (embedded in the Worker, never in dist/)
│   ├── worker/                 # Cloudflare Worker: data + image proxy, admin API, D1 migrations
│   ├── shared/                 # rules shared by the Worker and the admin app
│   ├── src-tauri/              # Tauri 2 desktop + Android shell (Rust)
│   └── static/                 # icons, social image, robots.txt, _headers
│
├── docs/                       # Legal + acknowledgements + contact
├── .github/workflows/          # data pipeline, discovery, CI, desktop/Android release
├── LICENSE                     # MIT — code
└── LICENSE-DATA                # CC BY 4.0 — dataset
Pipeline scripts
scripts/
├── core/
│   ├── constants.py        # Config, rate limits, anti-cheat patterns
│   ├── steam_client.py     # HTTP client, backoff, retry, session pooling
│   ├── data_store.py       # JSONL I/O, validation, dedup, save_main() + index.json
│   ├── overrides.py        # Human corrections re-applied on every write
│   ├── blocklist.py        # Removed games, left out of every write
│   ├── fetcher.py          # Batch processor, apply_details/reviews/players/scraped
│   ├── scraper.py          # Store page HTML parser (languages, DLC, tags, anti-cheat)
│   ├── anticheat.py        # Steam's anti-cheat disclosure and the rule that applies it
│   ├── steamcharts.py      # SteamCharts all-time peak reader (polite, stops on refusal)
│   ├── peaks.py            # all_time_peak: only ever raised
│   ├── rotation.py         # crc32 daily slices shared by the refresh jobs, growing with the catalogue
│   ├── budget.py           # A job's time budget: stop in time to save, warn before it runs out
│   └── health_checker.py   # Game status detection with cached API data
├── discover_new.py         # Propose new games into the /admin review queue
├── ingest_new.py           # Add queued games (health check + fetch from Steam)
├── remove_games.py         # Apply data/blocklist.jsonl now (runs before every ingest)
├── update_data.py          # Daily: fill in incomplete records
├── refresh_store_data.py   # Daily rotation: tags/languages weekly, store details monthly
├── refresh_anticheat.py    # Daily: every game's anti-cheat, as its developer declares it on Steam
├── refresh_peaks.py        # Daily rotation: all-time peak from SteamCharts, monthly per game
├── update_reviews.py       # Daily: every game's review summary from Steam, ~2 minutes
├── mark_dead_games.py      # Flag online games with no players
├── purge_unhealthy.py      # Daily: remove delisted/not-free/unreleased (store API screen + appdetails)
├── top_online.py / top_offline.py  # Leaderboards
├── edit_game.py            # Overrides from the command line
├── refetch_all.py          # Force re-fetch, in parts (manual only)
└── delete_game.py          # Remove games (blocklist) from the command line

Automated workflows

Workflow Trigger What it does
Auto Update JSON Daily 00:00 UTC Fills in incomplete records (API + HTML scrape)
Discover New Games Daily 05:00 UTC Propose new free games into the /admin review queue
Ingest New On scripts/temp_info.jsonl or data/blocklist.jsonl push, and every 3 h Apply removals, then add approved games
Top Online Sun/Wed/Fri 03:00 UTC Live player leaderboard
Top Offline 1st + 15th of month 05:00 UTC Leaderboard for single-player F2P concurrents
Update Reviews Daily 06:00 UTC Every game's review summary from Steam's store API
Refresh Store Data Daily 16:00 UTC Rotating re-fetch: tags, languages, DLC flag weekly; name, image, description, developer/publisher, release date, platforms, Metacritic, DRM monthly. Then every game's anti-cheat from Steam's disclosure (committed separately; manual run with only: anticheat does just that)
Refresh All-Time Peaks Daily 19:30 UTC all_time_peak from SteamCharts, every game once a month, plus our own samples; manual run with backfill asks about every game
Purge Unhealthy Daily 07:00 UTC Store API screen of every game; removal once appdetails agrees
Mark Dead Games Mon + Thu 04:30 UTC Online + >1y + 0 players ≥14d → is_dead=true + 💀 note
Daily Snapshot Daily 23:00 UTC Per-game players and reviews into data/snapshots/
Force Re-fetch Manual only Re-fetch games from scratch, one part (1/8 … 8/8) a run
Web CI / Python Tests On web/** / scripts/** change Type check, tests, build and output checks
Release Desktop / Android On desktop-v* / android-v* tag Build the apps into a draft release
Announce Discussion On release published Auto-create a GitHub Discussion in Announcements

The site itself is deployed by Cloudflare Workers Builds on every push to main, not by a workflow.

Contribute

New games arrive on their own: Discover New Games sweeps the Steam store daily into a review queue, and the maintainer approves them at /admin — by hand, or with an auto-approve rule that has a daily cap and is off until turned on. Discovery only proposes, and every approved game is checked again by the pipeline (free, a game, reachable) before it is published. Missing a free game a few days after it went free? Post its store link in Discussions › Ideas.

The companion browser extension, steam-f2p-extension, is archived and no longer a way in.

Issues are for reports, not submissions: Bug report, Feature request, Delete game and Feedback. All reviewed by hand.

The manual way. Fork, append store links to scripts/temp_info.jsonl, and open a pull request:

{"link": "730"}
{"link": "https://store.steampowered.com/app/570/"}

Accepts a full URL, short URL, or bare appid. A line is a link and nothing else: any other key is ignored, and the game is fetched from Steam. A correction to a published game (genre, anti-cheat, notes…) goes in a Discussion; the maintainer records it as an override.

See CONTRIBUTING.md for the full flow.

Why this exists

Unemployed, introvert max level, dropped out uni year 3, mooching off family. Needed a clean F2P list without digging Steam every time "hôm nay chơi free cái gì đây". So this repo was born — hobby project of a "barely competent" dev pumped by AI.

  • v1 was spaghetti.
  • v2.0 was organised spaghetti.
  • v2.1 was spaghetti with proper seasoning. It got a scraper.
  • v3.0 was spaghetti with a website on top, a desktop app, and bilingual menus.
  • v4.0 is the same spaghetti, rebuilt, with a review queue so it stops needing to be cooked by hand. Still a hobby.

Warning

Disclaimer. Quality / fun / safety: who knows. The maintainer only vouches for games personally played (safe column: y = yes, ? = dunno, n = nope). Problems with a game are between you, Valve, and the developer of that game — not the maintainer. See docs/DISCLAIMER.md for the specific accuracy caveats.

Docs

  • AUTHORS — maintainer + canonical handle map + AI-assistant disclosure.
  • CHANGELOG — what changed in each version.
  • CONTRIBUTING — issue templates, suggesting a game, manual fork.
  • PC spec (VI) — maintainer's dev hardware + test devices.
  • Dev environment (VI) — IDE, toolchains (Python 3.15 / Node 26 / TypeScript 7 / Rust stable / Tauri 2), workflow.
  • Tauri build — desktop build prerequisites + signing + auto-update.
  • DISCLAIMER — accuracy caveats, no-warranty, the broke-maintainer note.
  • Terms of Use — usage agreement, contributions, governing law.
  • EULA — plain-language commentary on both licences.
  • Privacy Policy — no accounts, no tracking cookies, no profile; what Cloudflare logs as the host, the one cookieless page-view count, and where the site is not served.
  • Deployment — how a push becomes a deploy, and how to verify or roll one back.
  • Admin setup — Cloudflare Access and the GitHub App behind /admin.
  • Acknowledgements — credits, AI assistants, third-party libraries.
  • Third-party dev tooling — dead-code / lint / audit tools + the pre-publish check suite.
  • Contact — where to find the maintainer.
  • CODE OF CONDUCT — be chill.
  • SECURITY — vulnerability reporting.
  • web/README.md — stack details, per-phase changelog, build instructions.

Code is MIT; the dataset is CC BY 4.0 with carve-outs for fields the project does not own (description, header_image). Keep the right licence with you if you fork. Star if you find a hidden gem ✨

Last auto-update: daily via GitHub Actions. Manual runs from the Actions tab.

S
Description
No description provided
Readme MIT
147 MiB
0 Stars 1 Watchers 0 Forks
Languages
TypeScript 51%
Python 23.1%
Svelte 23%
CSS 1%
JavaScript 0.7%
Other 1.2%