feat: guide AI agents to the fewest files

llms.txt and an ARD manifest name the pages and the JSON API, and lead
an agent setting up a device to --for-games --required-only. Every page
registers four read-only WebMCP tools: the setup command, a hash lookup,
an emulator's files and the site search.
This commit is contained in:
Abdessamad Derraz committed 2026-11-06 19:58:42 +01:00
1 parent 072380c268
commit 24d7db0ab3
8 files changed
+432 -3

No files matched your search

+1
View File
@@ -45,6 +45,7 @@ on:
- "scripts/restore_large_files.py"
- "scripts/safeparse.py"
- "scripts/sect233r1.py"
- "scripts/siteagents.py"
- "scripts/sitebootstrap.py"
- "scripts/sitediagrams.py"
- "scripts/siteentries.py"
+6 -2
View File
@@ -32,7 +32,7 @@ import urllib.parse
import urllib.request
import zipfile
from collections import Counter
from collections.abc import Callable, Iterator
from collections.abc import Callable, Iterable, Iterator
from datetime import datetime, timezone
from pathlib import Path
from typing import TypeVar
@@ -92,6 +92,7 @@ import sitemarkup as markup
import siteentries
import siteinputs
import retired_files
import siteagents
import sitestarlight
from siteexports import generate_exports_page, load_export_reports
from siteguides import (
@@ -8065,7 +8066,7 @@ def generate_emulators_index(profiles: dict) -> str:
return "\n".join(lines) + "\n"
def _snippet_noun(names, firmware: int, fetched: int) -> str:
def _snippet_noun(names: Iterable[str], firmware: int, fetched: int) -> str:
"""What an emulator's files are called in its snippet. A Switch
emulator reads keys, and "eden keys" is what readers search for; Ship
of Harkinian reads game ROMs, which are no BIOS."""
@@ -12528,6 +12529,9 @@ def _generate(args: argparse.Namespace) -> None:
stats = compute_stats(db, coverages, profiles)
stats["composition"] = compute_composition(db)
write_if_changed(str(public / "stats.json"), generate_stats(stats))
# llms.txt and the ARD manifest an agent reads first, from the registry.
siteagents.write(public, sorted(((name, cov["platform"]) for name, cov in coverages.items()),
key=lambda row: row[1].casefold()), write_if_changed)
# Computed once: the gap analysis page and the published gaps export must
# report the same files.
+157
View File
@@ -0,0 +1,157 @@
"""What the site tells an AI agent about itself.
- ``llms.txt`` (https://llmstxt.org): the pages and data an agent should
read first, in Markdown, at the site's root.
- ``ard.json``: an Agentic Resource Discovery manifest
(https://github.com/ards-project/ard-spec) listing that file and the JSON
API, linked from every page's head with ``rel="ard"``.
Both are written from the registry at each build, so a new platform or a
moved page cannot leave them stale. The WebMCP tools a page registers live
in website/src/scripts/agent-tools.ts.
"""
from __future__ import annotations
import json
from collections.abc import Callable
from pathlib import Path
SITE = "https://abdess.github.io/retrobios/"
PUBLISHER = "abdess.github.io"
SH = "curl -fsSL https://abdess.github.io/retrobios/i | sh -s --"
PS = "& ([scriptblock]::Create((irm https://abdess.github.io/retrobios/i.ps1)))"
START_PAGES = (
("Which pack do I need", "which-pack/", "pick a device or setup, get the folder its BIOS files go in, "
"one command that installs and checks them, or a ZIP pack"),
("First install", "wiki/first-install/", "the steps from nothing to a working BIOS folder"),
("What are BIOS files", "wiki/what-are-bios-files/", "why emulators need them and where they come from"),
("Troubleshooting", "wiki/troubleshooting/", "a file the emulator does not find or refuses"),
("Common questions", "wiki/faq/", "short answers"),
)
DATA = (
("API index", "api/v1/index.json", "every dataset of the JSON API, versioned"),
("Catalog", "api/v1/catalog.json", "each dataset, manifest, export and download with its size, SHA-256 and schema"),
("Emulators", "api/v1/emulators.json", "every emulator profile: the files it reads, required or not, "
"their hashes and the source line that reads each. One document "
"per emulator at api/v1/emulators/<id>.json"),
("Core names", "api/v1/cores.json", "each core name and alias, with the profile it designates"),
("Systems", "api/v1/systems.json", "each system id, its name and the section holding its files"),
("Platforms", "api/v1/platforms.json", "each setup (RetroArch, Batocera...), its BIOS folder and files"),
("Hash lookup", "api/v1/lookup/00.json", "what a hash (MD5, SHA-1, SHA-256, CRC32) designates, sharded "
"by its first two hex characters: api/v1/lookup/<xx>.json"),
)
MORE = (
("Emulators", "emulators/", "one page per emulator"),
("Systems", "systems/", "one page per console maker, every file with its hashes"),
("Installer", "wiki/installer/", "every option of install.py"),
("What we count", "wiki/privacy/", "the site's cookie-free counting"),
("Is it legal?", "wiki/legal/", "the project's position"),
("Contributing", "contributing/", "add a missing file or an emulator profile"),
)
def llms_txt(platforms: list[tuple[str, str]]) -> str:
"""The site's llms.txt. ``platforms`` lists (id, display name) of every
registered setup, each with its page."""
lines = [
"# RetroBIOS",
"",
"> BIOS and firmware files for emulators, each checked against the emulator's source "
"code: which files an emulator reads, whether it starts without them, where each one "
"goes on every setup, their hashes, and an installer that puts them in place.",
"",
"Every claim about a file cites the line of emulator source code that reads it. The "
"files themselves are served by the GitHub repository "
"(https://github.com/Abdess/retrobios) and its releases.",
"",
"## Setting up a device for its user",
"",
"Most people want to play now, with the fewest files. The installer does that in one "
"command: it finds the setup and its BIOS folder, reads the games on the card, and with "
"`--for-games --required-only` installs only the files those games cannot start "
"without. Without these two options it installs every file of the setup's list "
"(gigabytes), even with `--yes`.",
"",
f"1. Preview, nothing written: `{SH} --for-games --required-only --dry-run`",
f"2. Install, no questions: `{SH} --for-games --required-only --yes`",
f"3. On Windows (PowerShell): `{PS} --for-games --required-only --yes`",
"",
"- Run it on the device itself (its terminal or SSH), or on a computer with the device's "
"card plugged in. The installer recognises the handheld systems on the card, among them "
"ArkOS, dArkOS, ArchR and UnofficialOS (R36S and its clones), muOS, KNULLI, ROCKNIX, "
"AmberELEC, EmuELEC, Onion, spruceOS and MinUI. "
"The [Linux handhelds page](" + SITE + "platforms/linux-handhelds/) maps each device "
"to its system and folder.",
"- From a computer, if the card is not found, name its BIOS folder: `--dest <card>/bios` "
"(ArkOS and dArkOS: the `bios` folder of the EASYROMS part).",
"- A games folder somewhere else: `--for-games <folder>`, one subfolder per system.",
"- Later, to see what is missing or wrong: `--check`. Machine-readable progress: "
"`--events` (JSON lines, one per event, documented on the installer page).",
"- A game the installer cannot identify gets every BIOS of its system, and the run names it.",
"",
"## Start",
"",
*(f"- [{title}]({SITE}{path}): {what}" for title, path, what in START_PAGES),
"",
"## Setups",
"",
*(f"- [{name}]({SITE}platforms/{pid}/): its BIOS folder and files" for pid, name in platforms),
"",
"## Data",
"",
*(f"- [{title}]({SITE}{path}): {what}" for title, path, what in DATA),
"",
"## Optional",
"",
*(f"- [{title}]({SITE}{path}): {what}" for title, path, what in MORE),
]
return "\n".join(lines) + "\n"
def ard_manifest() -> dict:
"""The ARD manifest: the llms.txt and the JSON API, each with the
questions it answers."""
return {
"entries": [
{
"identifier": f"urn:air:{PUBLISHER}:retrobios:guide",
"displayName": "RetroBIOS guide for agents",
"type": "text/markdown",
"url": f"{SITE}llms.txt",
"description": "The pages and datasets of RetroBIOS: BIOS and firmware files for "
"emulators, checked against each emulator's source code.",
"representativeQueries": [
"install the minimum BIOS files for the games on my R36S running dArkOS",
"which BIOS files does mGBA need",
"where do PlayStation BIOS files go on Batocera",
"how do I install BIOS files for RetroArch",
],
"tags": ["bios", "firmware", "emulation", "retrogaming"],
},
{
"identifier": f"urn:air:{PUBLISHER}:retrobios:api",
"displayName": "RetroBIOS JSON API",
"type": "application/json",
"url": f"{SITE}api/v1/index.json",
"description": "Versioned JSON datasets: emulator profiles, platforms, systems, core "
"names and a hash lookup.",
"representativeQueries": [
"what file has the MD5 490f666e1afb15b7362b406ed1cea246",
"which files does the beetle_psx core read",
"list the systems Batocera supports",
],
"capabilities": ["HashLookup", "EmulatorFiles", "PlatformFiles"],
"tags": ["api", "json", "bios"],
},
]
}
def write(public: Path, platforms: list[tuple[str, str]],
write_if_changed: Callable[[str, str], object]) -> None:
write_if_changed(str(public / "llms.txt"), llms_txt(platforms))
write_if_changed(str(public / "ard.json"), json.dumps(ard_manifest(), indent=1) + "\n")
+93
View File
@@ -0,0 +1,93 @@
"""What the site tells an AI agent: llms.txt, the ARD manifest, the WebMCP tools.
An agent asked to set up a handheld reads these first. They must lead it to
the fewest files the games need, and every address they give must exist.
"""
from __future__ import annotations
import tests # noqa: F401 offline, scratch under tmp/, the suite's own cache
import re
import sys
import unittest
from pathlib import Path
REPO_ROOT = Path(__file__).resolve().parent.parent
sys.path.insert(0, str(REPO_ROOT / "scripts"))
import siteagents # noqa: E402
DIST = REPO_ROOT / "website" / "dist"
TOOLS = REPO_ROOT / "website" / "src" / "scripts" / "agent-tools.ts"
FOOTER = REPO_ROOT / "website" / "src" / "components" / "Footer.astro"
CONFIG = REPO_ROOT / "website" / "astro.config.mjs"
class TheGuideLeadsToTheFewestFiles(unittest.TestCase):
def setUp(self) -> None:
self.text = siteagents.llms_txt([("retroarch", "RetroArch"), ("batocera", "Batocera")])
def test_it_is_llms_txt(self) -> None:
self.assertTrue(self.text.startswith("# RetroBIOS\n\n> "))
self.assertIn("\n## Setting up a device for its user\n", self.text)
def test_every_command_installs_only_what_the_games_need(self) -> None:
commands = re.findall(r"`((?:curl|&) [^`]+)`", self.text)
self.assertGreaterEqual(len(commands), 3)
for command in commands:
with self.subTest(command=command):
self.assertIn("--for-games --required-only", command)
def test_each_setup_has_its_page(self) -> None:
self.assertIn("(https://abdess.github.io/retrobios/platforms/batocera/)", self.text)
@unittest.skipUnless((DIST / "llms.txt").is_file(), "website/dist is not built")
def test_every_address_of_the_built_guide_exists(self) -> None:
built = (DIST / "llms.txt").read_text(encoding="utf-8")
for url in re.findall(r"\((https://abdess\.github\.io/retrobios/[^)]*)\)", built):
with self.subTest(url=url):
target = DIST / url.removeprefix(siteagents.SITE)
self.assertTrue(target.is_file() or (target / "index.html").is_file(), url)
class TheManifestIsAnArdManifest(unittest.TestCase):
"""The terms ard-entry.schema.json requires (ards-project/ard-spec)."""
def test_entries(self) -> None:
entries = siteagents.ard_manifest()["entries"]
self.assertTrue(entries)
for entry in entries:
with self.subTest(entry=entry.get("identifier")):
self.assertRegex(entry["identifier"], r"^urn:air:[a-zA-Z0-9.-]+(:[a-zA-Z0-9._-]+)+$")
self.assertTrue(entry["displayName"])
self.assertRegex(entry["type"], r"^[a-z]+/[\w.+-]+$")
self.assertTrue(("url" in entry) != ("data" in entry))
self.assertTrue(2 <= len(entry["representativeQueries"]) <= 5)
self.assertTrue(entry["url"].startswith(siteagents.SITE))
def test_every_page_links_it(self) -> None:
self.assertIn("rel: 'ard'", CONFIG.read_text(encoding="utf-8"))
class TheToolsReadOnly(unittest.TestCase):
def setUp(self) -> None:
self.source = TOOLS.read_text(encoding="utf-8")
def test_every_page_registers_them(self) -> None:
self.assertIn("import '../scripts/agent-tools.ts';", FOOTER.read_text(encoding="utf-8"))
names = re.findall(r"^\s*name: '([a-z_]+)'", self.source, re.M)
self.assertEqual(names, ["lookup_bios_hash", "emulator_bios_files", "search_retrobios",
"device_setup_command"])
self.assertEqual(self.source.count("readOnlyHint: true"), len(names))
def test_the_setup_command_defaults_to_the_games(self) -> None:
self.assertIn("const games = everything ? [] :", self.source)
self.assertIn("'--required-only'", self.source)
def test_they_reach_no_other_site(self) -> None:
code = re.sub(r"^\s*//.*$", "", self.source, flags=re.M)
self.assertEqual(set(re.findall(r"https://[\w.-]+", code)), {"https://abdess.github.io"})
if __name__ == "__main__":
unittest.main()
+3
View File
@@ -77,6 +77,9 @@ export default defineConfig({
// Ownership of the URL-prefix property https://abdess.github.io/retrobios/
// in Google Search Console. Removing it unverifies the property.
{ tag: 'meta', attrs: { name: 'google-site-verification', content: '8SueST3hNHl8_vM-OJJP0rz-ZMlI2DALLpChrwZoGMs' } },
// What an AI agent reads first (scripts/siteagents.py): the ARD manifest lists
// llms.txt and the JSON API.
{ tag: 'link', attrs: { rel: 'ard', type: 'application/json', href: `${BASE}ard.json` } },
],
social: [{ icon: 'github', label: 'RetroBIOS on GitHub', href: 'https://github.com/Abdess/retrobios' }],
customCss: ['./src/styles/retrobios.css'],
+1
View File
@@ -48,6 +48,7 @@ const date = releaseDate(tag);
<script>
import '../scripts/count.ts';
import '../scripts/agent-tools.ts';
</script>
<style>
+170
View File
@@ -0,0 +1,170 @@
// WebMCP tools (https://developer.chrome.com/docs/ai/webmcp): what an AI
// agent browsing the site can ask it directly, instead of reading pages.
// Each tool reads the JSON the site already publishes (api/v1/) or its
// search index, changes nothing and sends nothing anywhere else. A browser
// without WebMCP skips the whole file.
type ToolResult = string;
type Tool = {
name: string;
description: string;
inputSchema: Record<string, unknown>;
annotations?: { readOnlyHint?: boolean };
execute: (input: Record<string, unknown>) => Promise<ToolResult>;
};
type ModelContext = { registerTool: (tool: Tool) => Promise<void> | void };
const BASE = import.meta.env.BASE_URL.replace(/\/?$/, '/');
const SITE = new URL(BASE, location.origin);
const json = async <T>(path: string): Promise<T> => {
const response = await fetch(new URL(path, SITE));
if (!response.ok) throw new Error(`${path}: HTTP ${response.status}`);
return response.json() as Promise<T>;
};
const absolute = (url: string) => new URL(url, location.origin).href;
type LookupItem = { algorithm: string; layer: string; name?: string; value: string; url: string; owner?: string; sha1?: string };
type ProfileFile = Record<string, unknown> & { name?: string; required?: boolean };
type EmulatorDoc = { items: { id: string; profile: { emulator?: string; systems?: string[]; files?: ProfileFile[] } }[] };
type CoreIndex = { items: { name: string; profile: string; preferred?: boolean }[] };
const HASH = /^(?:[0-9a-f]{8}|[0-9a-f]{32}|[0-9a-f]{40}|[0-9a-f]{64})$/;
const lookupHash: Tool = {
name: 'lookup_bios_hash',
description:
'Find which BIOS or firmware file a hash designates in the RetroBIOS collection, and the emulators or setups that '
+ 'declare it. Accepts a CRC32 (8 hex characters), MD5 (32), SHA-1 (40) or SHA-256 (64).',
inputSchema: {
type: 'object',
properties: { hash: { type: 'string', description: 'The hash, in hexadecimal.' } },
required: ['hash'],
},
annotations: { readOnlyHint: true },
async execute({ hash }) {
const value = String(hash ?? '').trim().toLowerCase().replace(/^0x/, '');
if (!HASH.test(value)) return 'Not a CRC32, MD5, SHA-1 or SHA-256 value.';
const shard = await json<{ items: LookupItem[] }>(`api/v1/lookup/${value.slice(0, 2)}.json`);
const found = shard.items.filter((item) => item.value === value);
if (!found.length) return `No file in the collection or its profiles has the ${value.length === 8 ? 'CRC32' : 'hash'} ${value}.`;
return JSON.stringify(found.map((item) => ({ ...item, url: absolute(item.url) })), null, 1);
},
};
const FILE_FIELDS = ['name', 'path', 'required', 'system', 'size', 'md5', 'sha1', 'crc32', 'sha256', 'region', 'note'];
const emulatorFiles: Tool = {
name: 'emulator_bios_files',
description:
'List the BIOS and firmware files an emulator or libretro core reads, from its profile checked against its source code: '
+ 'each file name, where it goes, whether the emulator starts without it, and its hashes.',
inputSchema: {
type: 'object',
properties: {
emulator: { type: 'string', description: 'The emulator or core name, for example "mGBA", "beetle_psx" or "DuckStation".' },
},
required: ['emulator'],
},
annotations: { readOnlyHint: true },
async execute({ emulator }) {
const asked = String(emulator ?? '').trim();
const key = asked.toLowerCase().replace(/[\s_]+/g, '_');
const cores = await json<CoreIndex>('api/v1/cores.json');
const named = cores.items.filter((core) => [core.name.toLowerCase(), core.profile].includes(key)
|| core.name.toLowerCase().replace(/[\s-]+/g, '_') === key);
const id = (named.find((core) => core.preferred) ?? named[0])?.profile ?? key.replace(/_/g, '-');
let doc: EmulatorDoc;
try {
doc = await json<EmulatorDoc>(`api/v1/emulators/${encodeURIComponent(id)}.json`);
} catch {
return `No emulator profile named "${asked}". The list is at ${absolute(`${BASE}emulators/`)}.`;
}
const { profile } = doc.items[0];
const files = (profile.files ?? []).map((file) =>
Object.fromEntries(FILE_FIELDS.filter((field) => file[field] !== undefined).map((field) => [field, file[field]])));
return JSON.stringify({
emulator: profile.emulator ?? id,
page: absolute(`${BASE}emulators/${id}/`),
systems: profile.systems ?? [],
files,
}, null, 1);
},
};
type Pagefind = { search: (query: string) => Promise<{ results: { data: () => Promise<{ url: string; meta: { title?: string }; excerpt: string }> }[] }> };
const searchSite: Tool = {
name: 'search_retrobios',
description: 'Search the RetroBIOS site: emulators, consoles, setups (RetroArch, Batocera, EmuDeck...) and guides. Returns the best pages.',
inputSchema: {
type: 'object',
properties: { query: { type: 'string', description: 'Words to search for, for example "dreamcast bios batocera".' } },
required: ['query'],
},
annotations: { readOnlyHint: true },
async execute({ query }) {
const words = String(query ?? '').trim();
if (!words) return 'Nothing to search for.';
const pagefind = (await import(/* @vite-ignore */ `${BASE}pagefind/pagefind.js`)) as Pagefind;
const { results } = await pagefind.search(words);
const top = await Promise.all(results.slice(0, 8).map((result) => result.data()));
if (!top.length) return `No page matches "${words}".`;
return JSON.stringify(top.map((page) => ({
title: page.meta.title ?? '',
url: absolute(page.url),
excerpt: page.excerpt.replace(/<[^>]+>/g, ''),
})), null, 1);
},
};
// The command an agent runs to set up its user's device, from the same
// options wiki/installer.md documents: the fewest files the games on the
// card need, unless the user asks for everything.
const SH = 'curl -fsSL https://abdess.github.io/retrobios/i | sh -s --';
const PS = '& ([scriptblock]::Create((irm https://abdess.github.io/retrobios/i.ps1)))';
const setupCommand: Tool = {
name: 'device_setup_command',
description:
'The one command that installs the BIOS files a device or computer needs, with RetroBIOS\'s installer. By default it '
+ 'installs only what the games on the card need to start, which is what most players want. It finds the setup '
+ '(RetroArch, Batocera, ArkOS, dArkOS, muOS, KNULLI, ROCKNIX...) and its BIOS folder itself.',
inputSchema: {
type: 'object',
properties: {
shell: { type: 'string', enum: ['sh', 'powershell'], description: '"sh" on Linux, macOS and handhelds (the default), "powershell" on Windows.' },
everything: { type: 'boolean', description: 'Install every file of the setup\'s list instead of only what the games need. Gigabytes.' },
bios_folder: { type: 'string', description: 'The BIOS folder, when run on a computer and the card is not found, for example "/media/user/EASYROMS/bios".' },
games_folder: { type: 'string', description: 'A games folder that is not on the card, one subfolder per system.' },
platform: { type: 'string', description: 'The setup, only when it is not detected (names from the installer\'s --list-platforms).' },
},
},
annotations: { readOnlyHint: true },
async execute({ shell, everything, bios_folder, games_folder, platform }) {
const quote = (value: unknown) => `'${String(value).replace(/'/g, shell === 'powershell' ? "''" : "'\\''")}'`;
const run = shell === 'powershell' ? PS : SH;
const place: string[] = [];
if (platform) place.push('--platform', quote(platform));
if (bios_folder) place.push('--dest', quote(bios_folder));
const games = everything ? [] : [...(games_folder ? ['--for-games', quote(games_folder)] : ['--for-games']), '--required-only'];
const command = (...more: string[]) => [run, ...place, ...more].join(' ');
return JSON.stringify({
preview: command(...games, '--dry-run'),
install: command(...games, '--yes'),
check_later: command('--check'),
notes: [
'Run the preview first: it writes nothing and lists what would be downloaded.',
'Run on the device itself, or on a computer with the device\'s card plugged in.',
'A game the installer cannot identify gets every BIOS of its system, and the run names it.',
`Guide: ${absolute(`${BASE}llms.txt`)}`,
],
}, null, 1);
},
};
const context = (document as Document & { modelContext?: ModelContext }).modelContext;
if (context?.registerTool) {
for (const tool of [setupCommand, lookupHash, emulatorFiles, searchSite]) {
Promise.resolve(context.registerTool(tool)).catch((error) => console.warn(`WebMCP: ${tool.name}:`, error));
}
}
+1 -1
View File
@@ -67,7 +67,7 @@ Every module in `tests/`, by the area it guards.
| Pack probe | the cores a platform runs for a system, staging grown to a fixed point, runs mapped to pack and manifest entries, the line and its schema, the probe lock per run, that no page shows its lines. One class builds a real pack narrowed to PlayStation and lays it out both ways, from the archive and with `install.py --dest`, and skips without the collection | `test_pack_probe` |
| Automation runner, disk cleanup, asset sync and bit-flip repair | fake jobs, git repositories with a bare origin and session worktrees, a committed file with one bit flipped, canned CI runs in place of `gh` | `test_autorun` |
| Truth and native export | synthetic + cached originals | `test_export_counts`, `test_export_native_layer`, `test_export_report`, `test_exporters`, `test_truth_data_dir_readers`, `test_truth_diff`, `test_truth_layout` |
| Scrapers, upstream and provenance | inline sources, synthetic | `test_batocera_md5_list`, `test_check_freshness`, `test_cited_lines`, `test_contributing_text`, `test_coverage_cache`, `test_data_dir_refs`, `test_dataclass_defaults`, `test_dedup`, `test_doc_pages`, `test_doc_surfaces`, `test_elec_scrapers`, `test_fbneo_parser`, `test_find_pinned_revision`, `test_hash_merge`, `test_launchers`, `test_launchers_emudeck`, `test_launchers_recalbox`, `test_launchers_retrobat`, `test_launchers_retrodeck`, `test_launchers_retropie`, `test_launchers_rocknix`, `test_mame_devices`, `test_mame_parser`, `test_mame_ref_audit`, `test_optional_dependencies`, `test_pin_build`, `test_profile_sync`, `test_prose_style`, `test_provenance`, `test_refresh_data_dirs`, `test_refresh_failures`, `test_refresh_stale`, `test_retrodeck_scraper_errors`, `test_retired_files`, `test_romset_recipes`, `test_schema_semantic_gate`, `test_scraper_contract`, `test_site_arcade_boards`, `test_site_availability`, `test_site_bootstrap`, `test_site_built_gate`, `test_site_components`, `test_site_console_folders`, `test_site_content_data`, `test_site_copy`, `test_site_cores`, `test_site_css`, `test_site_density_cards`, `test_site_emulator_page`, `test_site_emulators`, `test_site_entry`, `test_site_entry_tables`, `test_site_exports`, `test_site_funnel`, `test_site_images`, `test_site_inputs`, `test_site_insights`, `test_site_integration`, `test_site_nav`, `test_site_picker`, `test_site_platform_density`, `test_site_platform_hashes`, `test_site_privacy`, `test_site_reference`, `test_site_render`, `test_site_search`, `test_site_system_links`, `test_site_system_names`, `test_site_system_sections`, `test_site_systems`, `test_site_validation`, `test_site_weight`, `test_site_wiki`, `test_site_write`, `test_source_ref_targets`, `test_upstream`, `test_upstream_index`, `test_validate_pr`, `test_validate_pr_flow`, `test_validate_pr_hashes`, `test_version_basis`, `test_workflow_paths` |
| Scrapers, upstream and provenance | inline sources, synthetic | `test_batocera_md5_list`, `test_check_freshness`, `test_cited_lines`, `test_contributing_text`, `test_coverage_cache`, `test_data_dir_refs`, `test_dataclass_defaults`, `test_dedup`, `test_doc_pages`, `test_doc_surfaces`, `test_elec_scrapers`, `test_fbneo_parser`, `test_find_pinned_revision`, `test_hash_merge`, `test_launchers`, `test_launchers_emudeck`, `test_launchers_recalbox`, `test_launchers_retrobat`, `test_launchers_retrodeck`, `test_launchers_retropie`, `test_launchers_rocknix`, `test_mame_devices`, `test_mame_parser`, `test_mame_ref_audit`, `test_optional_dependencies`, `test_pin_build`, `test_profile_sync`, `test_prose_style`, `test_provenance`, `test_refresh_data_dirs`, `test_refresh_failures`, `test_refresh_stale`, `test_retrodeck_scraper_errors`, `test_retired_files`, `test_romset_recipes`, `test_schema_semantic_gate`, `test_scraper_contract`, `test_site_agents`, `test_site_arcade_boards`, `test_site_availability`, `test_site_bootstrap`, `test_site_built_gate`, `test_site_components`, `test_site_console_folders`, `test_site_content_data`, `test_site_copy`, `test_site_cores`, `test_site_css`, `test_site_density_cards`, `test_site_emulator_page`, `test_site_emulators`, `test_site_entry`, `test_site_entry_tables`, `test_site_exports`, `test_site_funnel`, `test_site_images`, `test_site_inputs`, `test_site_insights`, `test_site_integration`, `test_site_nav`, `test_site_picker`, `test_site_platform_density`, `test_site_platform_hashes`, `test_site_privacy`, `test_site_reference`, `test_site_render`, `test_site_search`, `test_site_system_links`, `test_site_system_names`, `test_site_system_sections`, `test_site_systems`, `test_site_validation`, `test_site_weight`, `test_site_wiki`, `test_site_write`, `test_source_ref_targets`, `test_upstream`, `test_upstream_index`, `test_validate_pr`, `test_validate_pr_flow`, `test_validate_pr_hashes`, `test_version_basis`, `test_workflow_paths` |
| Games on the card | small DATs, listxml and setup sources, a fake fetcher, fixture install lists, and the committed data against its schemas | `test_arcade_parents`, `test_game_catalogs`, `test_game_headers`, `test_game_regions`, `test_games_index`, `test_romfolders_card`, `test_romfolders_es` |
## Test architecture