Files
shopdb-flask/shopdb/core/api/docs.py
cproudlock 68567d9a27 Add a try-it API page, and stop the docs page reaching for the internet
/api/docs renders the spec with Redoc, which is read-only by design. The thing
people mean when they ask for Swagger is firing a request from the page, so
Swagger UI now sits beside it at /api/docs/try, reading the same spec - one
contract, two renderers.

It is OFF outside development and 404s rather than rendering a disabled page: a
production box holds live fleet data, and a docs page that can POST to it is a
different object from a docs page that cannot.

Vendored, not linked. swagger-ui-bundle.min.js and swagger-ui.css are served
from staticdocs/ beside the existing Redoc bundle, because production has no
internet. validatorUrl is null: left unset, Swagger UI hands the spec to the
public swagger validator to render a badge, which here is a request that cannot
succeed and elsewhere would ship the internal API surface off the box.

THE VENDORED REDOC BUNDLE WAS ALREADY REACHING THE INTERNET, found while
proving this page does not. redoc.standalone.js fetched
cdn.redoc.ly/redoc/logo-mini.svg for the "API docs by Redocly" mark, so every
load of /api/docs on the air-gapped server has been waiting on a request that
cannot complete. Replaced with an inline transparent GIF.

Three gate holes closed, each one a dodge that worked:

- the page is built by interpolating url_for output into a JS string literal,
  unescaped. SCRIPT_NAME is proxy-controlled and ProxyFix honours
  X-Forwarded-Prefix, so a mount path carrying a quote closed the literal and
  executed what followed. The spec URL is now json.dumps'd and the rest
  HTML-escaped; the injected quote renders as \" and is inert
- the "no internet" assertion on the page scanned for https?://, so
  <script src="//cdn.jsdelivr.net/..."> - the form a copy-paste from any CDN's
  own documentation produces - matched nothing and passed
- the same assertion on the vendored files was a five-host CDN allowlist. An
  upstream bump to any other host walked through it, and the one real leak in
  that directory was on no list. It now matches the SHAPES a browser fetches -
  @import, css url(), src/href, @font-face - whatever the host, ignoring the
  data: URIs this CSS is full of

And the Redoc page's link to the try-it view is built with url_for, which raises
BuildError when those routes are not mounted - every production box. Made
unconditional it would 500 the docs page in exactly the deployment nobody tests
locally, and no test caught that. One does now.
2026-09-18 20:18:23 -04:00

211 lines
9.2 KiB
Python

"""Interactive API docs: self-hosted Redoc over the generated OpenAPI spec.
Served at /api/docs (relative to the mount). The Redoc bundle is vendored in
staticdocs/ so this works fully offline on the air-gapped prod box - no CDN.
The spec is docs/openapi.json in the repo (regenerate with
scripts/gen_openapi.py after API changes).
Two renderers over ONE spec. Redoc is read-only by design and is the default,
because what is published is a contract. Swagger UI at /api/docs/try adds the
one thing Redoc will not do: fire a real request from the page, against this
server, with whatever session the browser is already carrying. That is its
whole value and the whole reason it is off unless a site turns it on - see
API_DOCS_TRYIT_ENABLED in shopdb/config.py.
"""
import json
import os
from markupsafe import escape
from flask import Blueprint, Response, current_app, send_file, url_for
docs_bp = Blueprint('docs', __name__)
_HERE = os.path.dirname(os.path.abspath(__file__))
# shopdb/core/api -> repo root
_REPO_ROOT = os.path.dirname(os.path.dirname(os.path.dirname(_HERE)))
_SPEC = os.path.join(_REPO_ROOT, 'docs', 'openapi.json')
_LLMS = os.path.join(_REPO_ROOT, 'docs', 'llms.txt')
_REDOC = os.path.join(_HERE, 'staticdocs', 'redoc.standalone.js')
# Vendored from swagger-ui-dist 5.33.0. TWO files out of the ~30 in that dist,
# because those are the two the page below actually loads:
# swagger-ui.css - the stylesheet. Its images are inline data: URIs,
# so it pulls nothing else. The trailing
# sourceMappingURL comment was stripped: the .map
# is not vendored, and left in place it turns every
# devtools open into a 404 on a box with no
# internet to fall back on.
# swagger-ui-bundle.min.js - the renderer plus the core presets, self-
# contained. NOT vendored beside it:
# swagger-ui-standalone-preset.js (the topbar with
# the "load any spec URL" box - this page serves
# one spec on purpose), the ES bundles (module
# builds of the same code), the .map files (~10 MB
# of source maps that only devtools reads),
# oauth2-redirect.html (no OAuth flow here), and
# the favicons.
# The upstream name is swagger-ui-bundle.js; the vendored copy carries .min.js,
# which is what it is - a minified single-line build. That suffix is also what
# scripts/check-naming-and-style.sh already filters vendored bundles by, so the
# style gate does not report thousands of hits in code nobody here wrote.
# swagger-ui.LICENSE.txt is the Apache-2.0 text that ships with the dist. It is
# never served; it is here because redistributing the bundle requires it.
_SWAGGER_CSS = os.path.join(_HERE, 'staticdocs', 'swagger-ui.css')
_SWAGGER_JS = os.path.join(_HERE, 'staticdocs', 'swagger-ui-bundle.min.js')
_TRYIT_ENDPOINT = 'docs.swagger_tryit'
def _tryit_is_mounted():
"""True when the try-it page exists on THIS app.
Reads the url map rather than the config flag on purpose: the routes are
bound once, at blueprint registration, so a config value changed after boot
would have the Redoc page advertise a link that answers 404.
"""
return _TRYIT_ENDPOINT in current_app.view_functions
@docs_bp.route('/', strict_slashes=False)
def docs_index():
"""Redoc page. url_for keeps the asset URLs correct under any mount."""
spec_url = url_for('docs.openapi_spec')
redoc_url = url_for('docs.redoc_js')
# Mentioned only when it is actually there. An unconditional note would be
# an advert for a 404 on every box that ships the shipped default.
tryit_note = ''
if _tryit_is_mounted():
tryit_note = (
'<p style="font:14px/1.4 system-ui,sans-serif;margin:0;'
'padding:10px 16px;background:#eef2f7;border-bottom:1px solid #cbd5e1">'
'This page is read-only. To send a real request against this '
'server, use <a href="%s">the try-it view</a>.</p>'
% url_for(_TRYIT_ENDPOINT)
)
page = (
'<!doctype html><html><head>'
'<title>ShopDB API</title>'
'<meta charset="utf-8">'
'<meta name="viewport" content="width=device-width, initial-scale=1">'
'<style>body{margin:0;padding:0}</style></head><body>'
'%s'
'<redoc spec-url="%s"></redoc>'
'<script src="%s"></script>'
'</body></html>' % (tryit_note, spec_url, redoc_url)
)
return Response(page, mimetype='text/html')
@docs_bp.route('/openapi.json')
def openapi_spec():
"""The generated OpenAPI 3.1 spec (machine + LLM readable)."""
if not os.path.isfile(_SPEC):
return Response('{"error":"openapi.json not generated"}',
status=404, mimetype='application/json')
return send_file(_SPEC, mimetype='application/json')
@docs_bp.route('/redoc.standalone.js')
def redoc_js():
"""Vendored Redoc bundle (offline)."""
return send_file(_REDOC, mimetype='application/javascript')
@docs_bp.route('/llms.txt')
def llms_txt():
"""Concise LLM-oriented API guide."""
if not os.path.isfile(_LLMS):
return Response('llms.txt not found', status=404, mimetype='text/plain')
return send_file(_LLMS, mimetype='text/plain')
def swagger_tryit():
"""Swagger UI over the SAME spec the Redoc page renders.
Every URL on the page comes from url_for, so the spec and both assets stay
same-origin and correct under a mount path (prod runs under /shopdb).
"""
spec_url = url_for('docs.openapi_spec')
css_url = url_for('docs.swagger_css')
js_url = url_for('docs.swagger_js')
redoc_page_url = url_for('docs.docs_index')
page = (
'<!doctype html><html><head>'
'<title>ShopDB API - try it</title>'
'<meta charset="utf-8">'
'<meta name="viewport" content="width=device-width, initial-scale=1">'
'<link rel="stylesheet" href="%(css)s">'
'<style>body{margin:0;padding:0}'
'.tryit-warn{font:14px/1.4 system-ui,sans-serif;background:#7f1d1d;'
'color:#fff;margin:0;padding:10px 16px}'
'.tryit-warn a{color:#fff}</style>'
'</head><body>'
'<p class="tryit-warn"><b>Execute sends a real request to this '
'server</b>, as whoever is signed in on this browser. A write '
'operation changes live data. The read-only contract is at '
'<a href="%(redoc)s">/api/docs</a>.</p>'
'<div id="swagger-ui"></div>'
'<script src="%(js)s"></script>'
'<script>\n'
'window.ui = SwaggerUIBundle({\n'
# json.dumps, NOT a bare "%s". SCRIPT_NAME is proxy-controlled and
# ProxyFix honours X-Forwarded-Prefix, so a mount path carrying a quote
# closed this string literal and executed what followed: SCRIPT_NAME of
# /a"+alert(1)+" rendered url: "/a"+alert(1)+"/api/docs/openapi.json".
# Escaping it as JSON is what makes the value data rather than code.
' url: %(spec)s,\n'
' dom_id: "#swagger-ui",\n'
' presets: [SwaggerUIBundle.presets.apis],\n'
' layout: "BaseLayout",\n'
' deepLinking: true,\n'
' filter: true,\n'
# 573 operations. Expanded by default, the page spends many seconds
# laying out before it answers a scroll.
' docExpansion: "none",\n'
' withCredentials: true,\n'
# THE air-gap setting. Left unset, Swagger UI renders a "valid" badge
# by handing this spec to the public swagger validator - an outbound
# request that cannot succeed here, and that would ship the internal
# API surface off the box if it ever could.
' validatorUrl: null\n'
'});\n'
'</script>'
'</body></html>'
% {'css': escape(css_url), 'js': escape(js_url),
'spec': json.dumps(spec_url),
'redoc': escape(redoc_page_url)}
)
return Response(page, mimetype='text/html')
def swagger_css():
"""Vendored Swagger UI stylesheet (offline)."""
return send_file(_SWAGGER_CSS, mimetype='text/css')
def swagger_js():
"""Vendored Swagger UI bundle (offline)."""
return send_file(_SWAGGER_JS, mimetype='application/javascript')
@docs_bp.record_once
def _mount_tryit(state):
"""Bind the try-it routes only when the flag is on for this app.
Binding at registration rather than checking the flag inside the view, so
that when it is off the routes DO NOT EXIST: the URL gets Flask's own 404
and confirms nothing, and the unauthenticated-GET sweep in
tests/test_core/test_public_read_surface.py never has to make an exception
for three routes that are only meant to be reachable on a dev box.
"""
if not state.app.config.get('API_DOCS_TRYIT_ENABLED'):
return
state.add_url_rule('/try', endpoint='swagger_tryit',
view_func=swagger_tryit, strict_slashes=False)
state.add_url_rule('/swagger-ui.css', endpoint='swagger_css',
view_func=swagger_css)
state.add_url_rule('/swagger-ui-bundle.min.js', endpoint='swagger_js',
view_func=swagger_js)