diff --git a/.github/workflows/compare.yaml b/.github/workflows/compare.yaml
new file mode 100644
index 00000000..d9d98038
--- /dev/null
+++ b/.github/workflows/compare.yaml
@@ -0,0 +1,20 @@
+name: Compare spec to docs
+on:
+  workflow_dispatch:
+jobs:
+  compare:
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+      - name: Install deps
+        working-directory: ./_docmatch
+        run: pip install .
+      - name: Fetch docs
+        working-directory: ./_docmatch
+        run: curl -L https://spoonacular.com/food-api/docs > docs.html
+      - name: Compare
+        working-directory: ./_docmatch
+        run: python src/docmatch.py
diff --git a/_docmatch/.gitignore b/_docmatch/.gitignore
new file mode 100644
index 00000000..16c37027
--- /dev/null
+++ b/_docmatch/.gitignore
@@ -0,0 +1,70 @@
+docs.html
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm-project.org/#use-with-ide
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
diff --git a/_docmatch/README.md b/_docmatch/README.md
new file mode 100644
index 00000000..ba293903
--- /dev/null
+++ b/_docmatch/README.md
@@ -0,0 +1,14 @@
+# docmatch
+
+This is a script that compares the spec with the HTML docs.
+
+Using [pdm](https://pdm-project.org/) for dependencies:
+
+```shell
+curl -L https://spoonacular.com/food-api/docs > docs.html
+pdm install
+pdm run src/docmatch.py
+```
+
+(just pip can be used for deps too, `pip install .`,
+with a manually managed virtualenv or in an otherwise isolated environment like a container)
diff --git a/_docmatch/pdm.lock b/_docmatch/pdm.lock
new file mode 100644
index 00000000..9a316ab9
--- /dev/null
+++ b/_docmatch/pdm.lock
@@ -0,0 +1,64 @@
+# This file is @generated by PDM.
+# It is not intended for manual editing.
+
+[metadata]
+groups = ["default", "dev"]
+strategy = ["cross_platform", "inherit_metadata"]
+lock_version = "4.4.1"
+content_hash = "sha256:73271ac30340365272424743c809342bbbed0255cbc87da4d053a6e366e73874"
+
+[[package]]
+name = "beautifulsoup4"
+version = "4.12.3"
+requires_python = ">=3.6.0"
+summary = "Screen-scraping library"
+groups = ["default"]
+dependencies = [
+    "soupsieve>1.2",
+]
+files = [
+    {file = "beautifulsoup4-4.12.3-py3-none-any.whl", hash = "sha256:b80878c9f40111313e55da8ba20bdba06d8fa3969fc68304167741bbf9e082ed"},
+    {file = "beautifulsoup4-4.12.3.tar.gz", hash = "sha256:74e3d1928edc070d21748185c46e3fb33490f22f52a3addee9aee0f4f7781051"},
+]
+
+[[package]]
+name = "genson"
+version = "1.2.2"
+summary = "GenSON is a powerful, user-friendly JSON Schema generator."
+groups = ["default"]
+files = [
+    {file = "genson-1.2.2.tar.gz", hash = "sha256:8caf69aa10af7aee0e1a1351d1d06801f4696e005f06cedef438635384346a16"},
+]
+
+[[package]]
+name = "json-repair"
+version = "0.17.1"
+requires_python = ">=3.7"
+summary = "A package to repair broken json strings"
+groups = ["default"]
+files = [
+    {file = "json_repair-0.17.1-py3-none-any.whl", hash = "sha256:e649d7b7f35a7ecfcef795a460b033dd4cd542dd8637235e8918f3dbee9f8827"},
+    {file = "json_repair-0.17.1.tar.gz", hash = "sha256:d32768d3397dc9256b3cabdc731812c11ac60e6baa2f50df3ad8b5f006363ede"},
+]
+
+[[package]]
+name = "jsonref"
+version = "1.1.0"
+requires_python = ">=3.7"
+summary = "jsonref is a library for automatic dereferencing of JSON Reference objects for Python."
+groups = ["default"]
+files = [
+    {file = "jsonref-1.1.0-py3-none-any.whl", hash = "sha256:590dc7773df6c21cbf948b5dac07a72a251db28b0238ceecce0a2abfa8ec30a9"},
+    {file = "jsonref-1.1.0.tar.gz", hash = "sha256:32fe8e1d85af0fdefbebce950af85590b22b60f9e95443176adbde4e1ecea552"},
+]
+
+[[package]]
+name = "soupsieve"
+version = "2.5"
+requires_python = ">=3.8"
+summary = "A modern CSS selector implementation for Beautiful Soup."
+groups = ["default"]
+files = [
+    {file = "soupsieve-2.5-py3-none-any.whl", hash = "sha256:eaa337ff55a1579b6549dc679565eac1e3d000563bcb1c8ab0d0fefbc0c2cdc7"},
+    {file = "soupsieve-2.5.tar.gz", hash = "sha256:5663d5a7b3bfaeee0bc4372e7fc48f9cff4940b3eec54a6451cc5299f1097690"},
+]
diff --git a/_docmatch/pyproject.toml b/_docmatch/pyproject.toml
new file mode 100644
index 00000000..99344465
--- /dev/null
+++ b/_docmatch/pyproject.toml
@@ -0,0 +1,19 @@
+[project]
+name = "docmatch"
+version = "0.1.0"
+description = "Matches the spoonacular API docs to the OpenAPI spec"
+authors = [
+    {name = "Val Packett", email = "val@packett.cool"},
+]
+dependencies = [
+    "beautifulsoup4>=4.12.3",
+    "genson>=1.2.2",
+    "json-repair>=0.17.1",
+    "jsonref>=1.1.0",
+]
+requires-python = ">=3.10.0"
+readme = "README.md"
+license = {text = "MIT"}
+
+[tool.pdm]
+distribution = false
diff --git a/_docmatch/src/docmatch.py b/_docmatch/src/docmatch.py
new file mode 100644
index 00000000..916e6688
--- /dev/null
+++ b/_docmatch/src/docmatch.py
@@ -0,0 +1,214 @@
+import json
+import jsonref
+import json_repair
+import bs4
+import sys
+import re
+from collections import defaultdict
+from pprint import pprint
+from genson import SchemaBuilder
+from typing import List, Dict, Optional
+from dataclasses import dataclass
+
+def eprint(*args, **kwargs):
+    print(*args, file=sys.stderr, **kwargs)
+
+# https://github.com/n-takumasa/json-with-comments/blob/main/jsonc/_util.py
+_REMOVE_C_COMMENT = r"""
+    ( # String Literal
+        \"(?:\\.|[^\\\"])*?\"
+    )
+    |
+    ( # Comment
+        \/\*.*?\*\/
+        |
+        \/\/[^\r\n]*?(?:[\r\n])
+    )
+    """
+
+def _remove_c_comment(text: str) -> str:
+    if text[-1] != "\n":
+        text = text + "\n"
+    return re.sub(
+        _REMOVE_C_COMMENT,
+        lambda x: x.group(1),
+        text,
+        flags=re.DOTALL | re.VERBOSE,
+    )
+
+def t(el: bs4.Tag | bs4.NavigableString | None) -> bs4.Tag:
+    assert isinstance(el, bs4.Tag)
+    return el
+
+def json2schema(els: List[str]) -> dict:
+    builder = SchemaBuilder()
+    for el in els:
+        el = _remove_c_comment(el.replace("...\n", "\n"))
+        try:
+            # builder.add_object(pyjson5.decode(el))
+            builder.add_object(json_repair.loads(el))
+        except Exception as e:
+            eprint('JSON failed to parse', el, e)
+    schema = builder.to_schema()
+    del schema['$schema']
+    return schema
+
+@dataclass
+class Param:
+    description: str
+    where: str
+    typ: str
+    example: str
+    required: bool
+
+@dataclass
+class Doc:
+    method: str
+    url: str
+    summary: str
+    description: str
+    params: Dict[str, Param]
+    req_schema: Optional[dict]
+    resp_schema: dict
+
+def parsedoc(doc: bs4.Tag):
+    if doc.find('div', class_='api-method') is None:
+        return
+    method = t(doc.find('div', class_='api-method')).get_text()
+    url = t(doc.find('div', class_='api-path')).get_text().replace('https://api.spoonacular.com', '')
+    params = {}
+    tbl = doc.find('table', class_='api-parameter-table')
+    for param in (t(tbl).find_all('tr') if tbl else []):
+        tds = param.find_all('td')
+        if len(tds) == 0:
+            continue
+        [par, typ, exa, des] = tds
+        req = not (par.find('b') is None)
+        par = t(par.find('code'))
+        par_where = [c for c in par['class'] if c.startswith('api-in-')][0]
+        par = par.get_text()
+        typ = typ.get_text()
+        exa = exa.get_text()
+        if typ == 'number':
+            exa = float(exa)
+        if typ == 'boolean':
+            exa = exa == 'true'
+        params[par] = Param(
+            description=des.get_text(),
+            where=par_where.replace('api-in-', ''),
+            typ=typ,
+            example=exa,
+            required=req,
+        )
+    examples = {}
+    for hdr in doc.find_all('div', class_='apiPathHeader'):
+        ex = []
+        for sib in hdr.find_next_siblings():
+            if not isinstance(sib, bs4.Tag):
+                continue
+            if sib.name == 'div' and 'apiPathHeader' in sib['class']:
+                break
+            if sib.name == 'pre' and 'language-json' in sib['class']:
+                ex.append(sib.get_text())
+        if len(ex) > 0:
+            examples[hdr.get_text()] = ex
+    examples_req = []
+    examples_res = []
+    if '(post body)' in params:
+        examples_req.append(params['(post body)'].example)
+        del params['(post body)']
+    for hdr, exs in examples.items():
+        if hdr == 'Example Response' or hdr == 'Example Request and Response':
+            examples_res += exs
+        elif hdr == 'Example Request' or hdr == 'Example Request Body' or hdr == 'Example Request and Body':
+            examples_req += exs
+        else:
+            eprint('Unknown apiPathHeader:', hdr)
+    # pprint(params)
+    # print('Req:')
+    # pprint(json2schema(examples_req))
+    # print('Res:')
+    ret = Doc(
+        summary=str(doc['jss-title']),
+        description=t(doc.find('p')).get_text(strip=True),
+        method=method,
+        url=url,
+        params=params,
+        req_schema=None,
+        resp_schema=json2schema(examples_res)
+    )
+    if len(examples_req) > 0:
+        ret.req_schema = json2schema(examples_req)
+    return ret
+
+def compare_schema(location, l_name, l, r_name, r):
+    if l.get('type') != r.get('type'):
+        if (l.get('type') == 'integer' and r.get('type') == 'number') \
+            or (r.get('type') == 'integer' and l.get('type') == 'number') \
+            or (l.get('type') == 'string' and r.get('type') == 'text') \
+            or (r.get('type') == 'string' and l.get('type') == 'text'):
+            return
+        print(f"** ! `{location}`: type mismatch: {l_name} `{l.get('type')}` vs {r_name} `{r.get('type')}`")
+        return
+    l_props = l.get('properties', {})
+    r_props = r.get('properties', {})
+    for k in r_props.keys():
+        if k not in l_props:
+            print(f"** ! `{location}`: property `{k}` present in {r_name} but not in {l_name}")
+    for k, v in l_props.items():
+        if k not in r_props:
+            print(f"** ! `{location}`: property `{k}` present in {l_name} but not in {r_name}")
+            continue
+        compare_schema(location + '.' + k, l_name, v, r_name, r_props[k])
+    if 'items' in l and 'items' in r:
+        compare_schema(location + '[]', l_name, l['items'], r_name, r['items'])
+
+with open("../spoonacular-openapi-3.json") as f:
+    spec = jsonref.load(f)
+
+with open("docs.html") as f:
+    soup = bs4.BeautifulSoup(f, 'html.parser')
+    docs = soup.find_all('section', class_='jss-doc')
+    for doc in docs:
+        pdoc = parsedoc(doc)
+        if not pdoc: continue
+        if not pdoc.url in spec['paths']:
+            print(f"* ! `{pdoc.url}` not found in spec")
+            continue
+        scm = spec['paths'][pdoc.url]
+        if not pdoc.method.lower() in scm:
+            print(f"* ! method `{pdoc.method.lower()}` not found in spec for `{pdoc.url}`")
+            continue
+        scm = scm[pdoc.method.lower()]
+        print(f"* `{pdoc.method.lower()}` `{pdoc.url}`")
+
+        spec_params = {}
+        for p in scm.get('parameters', []):
+            if not 'name' in p:
+                print(f"** ! unresolved/unnamed param in spec: {p}")
+                continue
+            spec_params[p['name']] = p
+        for name in spec_params.keys():
+            if not name in pdoc.params:
+                print(f"** ! param `{name}` found in spec, but not in doc")
+        for name, doc_param in pdoc.params.items():
+            if not name in spec_params:
+                print(f"** ! param `{name}` found in doc, but not in spec")
+                continue
+            spec_param = spec_params[name]
+            def print_cmp(name, par, spec, doc):
+                if (spec == 'integer' and doc == 'number') or (spec == 'string' and doc == 'text'):
+                    return
+                if spec != doc:
+                    print(f"** ! param `{name}`.`{par}`: spec `{spec}` vs doc `{doc}`")
+            print_cmp(name, 'required', spec_param.get('required'), doc_param.required)
+            print_cmp(name, 'type', spec_param.get('schema', {}).get('type'), doc_param.typ)
+            print_cmp(name, 'in', spec_param.get('in'), doc_param.where)
+
+        spec_req_sch = scm.get('requestBody', {}).get('content', {}).get('application/json', {}).get('schema')
+        if spec_req_sch and not pdoc.req_schema:
+            print(f"** request body schema found in spec, but could not infer one from docs")
+        if not spec_req_sch and pdoc.req_schema:
+            print(f"** ! request body schema not found in spec, but could infer one from docs")
+        if spec_req_sch and pdoc.req_schema:
+            compare_schema('(req body)', 'spec', spec_req_sch, 'docs-inferred', pdoc.req_schema)