diff --git a/CHANGELOG.md b/CHANGELOG.md
index 87c9368..2d4d918 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -13,6 +13,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Use all rewriting stuff from zimscraperlib
 - Remove most HTML / CSS / JS rewriting logic which is now part of zimscraperlib 5
 - Fix wombat setup settings (especially `isSW`) (#293)
+- Fork cdxj_indexer codebase (#428)
 
 ### Fixed
 
diff --git a/docs/software_architecture.md b/docs/software_architecture.md
new file mode 100644
index 0000000..c79e1d8
--- /dev/null
+++ b/docs/software_architecture.md
@@ -0,0 +1,26 @@
+# Software architecture
+
+## cdxj_indexer and warcio
+
+[cdxj_indexer Python library](https://pypi.org/project/cdxj-indexer/) is a thin wrapper over [warcio Python library](https://pypi.org/project/warcio/). It used to iterate all record in WARCs.
+
+It provide two main features:
+
+- Loop over several WARCs in a directory (A visit of a website may be stored in several WARCs in the same directory).
+- Provide a buffered access to warcs content (and not a "stream" (fileio) only api) (but monkey patching returned WarcRecord.
+
+Except that, scraper directly uses WarcRecord (returned by cdxj_indexer, implemented in warcio) to access metadata and such.
+
+cdxj_indexer usefull methods are currently forked in warc2zim, see https://github.com/openzim/warc2zim/pull/428 for details.
+
+## zimscraperlib
+
+[zimscraperlib Python library](https://pypi.org/project/zimscraperlib) is used for ZIM operations and for all HTML / CSS / JS rewriting operations (mostly around URL manipulations, but not only).
+
+## requests
+
+[requests Python library](https://pypi.org/project/requests/) is used to retrieve the custom CSS file when a URL is passed.
+
+## brotlipy
+
+[brotlipy Python library](https://pypi.org/project/brotlipy/) is used to access brotli content in WARC records (not part of warcio because it is an optional dependency).
diff --git a/pyproject.toml b/pyproject.toml
index 7609a2d..4c0d458 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -14,11 +14,11 @@ dependencies = [
   "jinja2==3.1.4", # also update version in build-system above and in build_js.sh
   # to support possible brotli content in warcs, must be added separately
   "brotlipy==0.7.0",
-  "cdxj_indexer==1.4.6",
   "tinycss2==1.4.0",
   "beautifulsoup4==4.12.3", # used to parse base href
   "lxml==5.3.0", # used by beautifulsoup4 for parsing html
   "python-dateutil==2.9.0.post0",
+  "multipart==1.2.1",
 ]
 dynamic = ["authors", "classifiers", "keywords", "license", "version", "urls"]
 
diff --git a/src/warc2zim/cdxj_indexer/__init__.py b/src/warc2zim/cdxj_indexer/__init__.py
new file mode 100644
index 0000000..14245da
--- /dev/null
+++ b/src/warc2zim/cdxj_indexer/__init__.py
@@ -0,0 +1,9 @@
+from warc2zim.cdxj_indexer.bufferiter import buffering_record_iter
+from warc2zim.cdxj_indexer.main import iter_file_or_dir
+from warc2zim.cdxj_indexer.postquery import append_method_query_from_req_resp
+
+__all__ = [
+    "append_method_query_from_req_resp",
+    "buffering_record_iter",
+    "iter_file_or_dir",
+]
diff --git a/src/warc2zim/cdxj_indexer/bufferiter.py b/src/warc2zim/cdxj_indexer/bufferiter.py
new file mode 100644
index 0000000..9407e31
--- /dev/null
+++ b/src/warc2zim/cdxj_indexer/bufferiter.py
@@ -0,0 +1,130 @@
+import hashlib
+import shutil
+import tempfile
+
+from warc2zim.cdxj_indexer.postquery import append_method_query_from_req_resp
+
+BUFF_SIZE = 1024 * 64
+
+
+# ============================================================================
+def buffering_record_iter(
+    record_iter, digest_reader=None, url_key_func=None, *, post_append=False
+):
+    prev_record = None
+
+    for record in record_iter:
+        buffer_record_content(record)
+
+        record.file_offset = record_iter.get_record_offset()
+        record.file_length = record_iter.get_record_length()
+
+        if digest_reader:
+            curr = digest_reader.tell()
+            digest_reader.seek(record.file_offset)
+            record_digest, digest_length = digest_block(
+                digest_reader, record.file_length
+            )
+            digest_reader.seek(curr)
+
+            if digest_length != record.file_length:
+                raise Exception(
+                    f"Digest block mismatch, expected {record.file_length}, "
+                    f"got {digest_length}"
+                )
+
+            record.record_digest = record_digest
+
+        req, resp = concur_req_resp(prev_record, record)
+
+        if not req or not resp:
+            if prev_record:
+                yield prev_record
+                prev_record.buffered_stream.close()
+            prev_record = record
+            continue
+
+        join_req_resp(req, resp, post_append, url_key_func)
+
+        yield prev_record
+        if prev_record:
+            prev_record.buffered_stream.close()
+        yield record
+        record.buffered_stream.close()
+        prev_record = None
+
+    if prev_record:
+        yield prev_record
+        prev_record.buffered_stream.close()
+
+
+# ============================================================================
+def concur_req_resp(rec_1, rec_2):
+    if not rec_1 or not rec_2:
+        return None, None
+
+    if rec_1.rec_headers.get_header("WARC-Target-URI") != rec_2.rec_headers.get_header(
+        "WARC-Target-URI"
+    ):
+        return None, None
+
+    if rec_2.rec_headers.get_header(
+        "WARC-Concurrent-To"
+    ) != rec_1.rec_headers.get_header("WARC-Record-ID"):
+        return None, None
+
+    if rec_1.rec_type == "response" and rec_2.rec_type == "request":
+        req = rec_2
+        resp = rec_1
+
+    elif rec_1.rec_type == "request" and rec_2.rec_type == "response":
+        req = rec_1
+        resp = rec_2
+
+    else:
+        return None, None
+
+    return req, resp
+
+
+# ============================================================================
+def buffer_record_content(record):
+    spool = tempfile.SpooledTemporaryFile(BUFF_SIZE)
+    shutil.copyfileobj(record.content_stream(), spool)
+    spool.seek(0)
+    record.buffered_stream = spool
+
+
+# ============================================================================
+def join_req_resp(req, resp, post_append, url_key_func=None):
+    if req.http_headers is None:
+        return
+
+    resp.req = req
+
+    method = req.http_headers.protocol
+    if post_append and method.upper() in ("POST", "PUT"):
+        url = req.rec_headers.get_header("WARC-Target-URI")
+        query, append_str = append_method_query_from_req_resp(req)
+        resp.method = method.upper()
+        resp.requestBody = query
+        resp.urlkey = url + append_str
+        if url_key_func:
+            resp.urlkey = url_key_func(resp.urlkey)
+        req.urlkey = resp.urlkey
+
+
+# ============================================================================
+def digest_block(reader, length):
+    count = 0
+    hasher = hashlib.sha256()
+
+    while length > 0:
+        buff = reader.read(min(BUFF_SIZE, length))
+        if not buff:
+            break
+        hasher.update(buff)
+        length -= len(buff)
+        count += len(buff)
+
+    return "sha256:" + hasher.hexdigest(), count
diff --git a/src/warc2zim/cdxj_indexer/main.py b/src/warc2zim/cdxj_indexer/main.py
new file mode 100644
index 0000000..8f9fab2
--- /dev/null
+++ b/src/warc2zim/cdxj_indexer/main.py
@@ -0,0 +1,17 @@
+import os
+
+ALLOWED_EXT = (".arc", ".arc.gz", ".warc", ".warc.gz")
+
+
+# =================================================================
+def iter_file_or_dir(inputs: list[str]):
+    for input_ in inputs:
+        if not isinstance(input_, str) or not os.path.isdir(input_):
+            yield input_
+            continue
+
+        for root, _, files in os.walk(input_):
+            for filename in files:
+                if filename.endswith(ALLOWED_EXT):
+                    full_path = os.path.join(root, filename)
+                    yield full_path
diff --git a/src/warc2zim/cdxj_indexer/postquery.py b/src/warc2zim/cdxj_indexer/postquery.py
new file mode 100644
index 0000000..2142b14
--- /dev/null
+++ b/src/warc2zim/cdxj_indexer/postquery.py
@@ -0,0 +1,173 @@
+import base64
+import json
+import sys
+from io import BytesIO
+from urllib.parse import unquote_plus, urlencode
+
+from multipart import MultipartParser
+
+MAX_QUERY_LENGTH = 4096
+
+
+# ============================================================================
+def append_method_query_from_req_resp(req):
+    len_ = req.http_headers.get_header("Content-Length")
+    content_type = req.http_headers.get_header("Content-Type")
+    stream = req.buffered_stream
+    stream.seek(0)
+
+    url = req.rec_headers.get_header("WARC-Target-URI")
+    method = req.http_headers.protocol
+    return append_method_query(method, content_type, len_, stream, url)
+
+
+# ============================================================================
+def append_method_query(method, content_type, len_, stream, url):
+    # if method == 'GET':
+    #    return '', ''
+
+    if method in ("POST", "PUT"):
+        query = query_extract(content_type, len_, stream)
+    else:
+        query = ""
+
+    if "?" not in url:
+        append_str = "?"
+    else:
+        append_str = "&"
+
+    append_str += "__wb_method=" + method
+    if query:
+        append_str += "&" + query
+
+    return query, append_str
+
+
+# ============================================================================
+def query_extract(mime, length, stream):
+    """
+    Extract a url-encoded form POST/PUT from stream
+    content length, return None
+    Attempt to decode application/x-www-form-urlencoded or multipart/*,
+    otherwise read whole block and b64encode
+    """
+    query_data = b""
+
+    try:
+        length = int(length)
+    except (ValueError, TypeError):
+        if length is None:
+            length = 8192
+        else:
+            return
+
+    while length > 0:
+        buff = stream.read(length)
+
+        length -= len(buff)
+
+        if not buff:
+            break
+
+        query_data += buff
+
+    if not mime:
+        mime = ""
+
+    query = ""
+
+    def handle_binary(query_data):
+        return f"__wb_post_data={ base64.b64encode(query_data).decode()}"
+
+    if mime.startswith("application/x-www-form-urlencoded"):
+        try:
+            query = unquote_plus(query_data.decode("utf-8"))
+        except UnicodeDecodeError:
+            query = handle_binary(query_data)
+
+    elif mime.startswith("multipart/"):
+        try:
+            boundary = mime.split("boundary=")[1]
+            parser = MultipartParser(BytesIO(query_data), boundary, charset="utf8")
+        except (ValueError, IndexError):
+            # Content-Type multipart/form-data may lack "boundary" info
+            query = handle_binary(query_data)
+        else:
+            values = []
+            for part in parser:
+                if part is None:
+                    continue
+                values.append((part.name, part.value))
+
+            query = urlencode(values, True)
+
+    elif mime.startswith("application/json"):
+        try:
+            query = json_parse(query_data)
+        except Exception:
+            if query_data:
+                try:
+                    sys.stderr.write(
+                        "Error parsing: " + query_data.decode("utf-8") + "\n"
+                    )
+                except Exception:  # noqa: S110 # nosec B110
+                    pass
+
+            query = ""
+
+    elif mime.startswith("text/plain"):
+        try:
+            query = json_parse(query_data)
+        except Exception:
+            query = handle_binary(query_data)
+
+    # Remove AMF parsing, we do not really need it in warc2zim and AMF library is not
+    # maintained at all
+    # elif mime.startswith("application/x-amf"):
+    #     query = amf_parse(query_data)
+
+    else:
+        query = handle_binary(query_data)
+
+    if query:
+        query = query[:MAX_QUERY_LENGTH]
+
+    return query
+
+
+def json_parse(string):
+    data = {}
+    dupes = {}
+
+    def get_key(n):
+        if n not in data:
+            return n
+
+        if n not in dupes:
+            dupes[n] = 1
+
+        dupes[n] += 1
+        return n + "." + str(dupes[n]) + "_"
+
+    def _parser(json_obj, name=""):
+        if isinstance(json_obj, dict):
+            for n, v in json_obj.items():
+                _parser(v, n)
+
+        elif isinstance(json_obj, list):
+            for v in json_obj:
+                _parser(v, name)
+
+        elif name:
+            data[get_key(name)] = str(json_obj)
+
+    try:
+        _parser(json.loads(string))
+    except json.decoder.JSONDecodeError:
+        if b"\n" in string:
+            for line in string.split(b"\n"):
+                _parser(json.loads(line))
+        else:
+            raise
+
+    return urlencode(data)
diff --git a/src/warc2zim/converter.py b/src/warc2zim/converter.py
index 04bfa46..a82bae4 100644
--- a/src/warc2zim/converter.py
+++ b/src/warc2zim/converter.py
@@ -34,7 +34,6 @@
 
 # from zimscraperlib import getLogger
 from bs4 import BeautifulSoup
-from cdxj_indexer import buffering_record_iter, iter_file_or_dir
 from dateutil import parser
 from jinja2 import Environment, PackageLoader
 from warcio import ArchiveIterator
@@ -51,6 +50,7 @@
 from zimscraperlib.zim import metadata
 from zimscraperlib.zim.creator import Creator
 
+from warc2zim.cdxj_indexer import buffering_record_iter, iter_file_or_dir
 from warc2zim.constants import logger
 from warc2zim.icon_finder import Icon, get_sorted_icons, icons_in_html
 from warc2zim.items import StaticArticle, StaticFile, WARCPayloadItem
diff --git a/tests/cdxj_indexer/test_postappend.py b/tests/cdxj_indexer/test_postappend.py
new file mode 100644
index 0000000..6bcc35e
--- /dev/null
+++ b/tests/cdxj_indexer/test_postappend.py
@@ -0,0 +1,203 @@
+from io import BytesIO
+
+from warc2zim.cdxj_indexer.postquery import append_method_query
+
+
+# ============================================================================
+class MethodQueryCanonicalizer:
+    def __init__(self, method, content_type, req_len, req_stream):
+        self.method = method
+        self.content_type = content_type
+        self.req_len = req_len
+        self.req_stream = req_stream
+
+    def append_query(self, url):
+        self.req_stream.seek(0)
+        query_only, full_string = append_method_query(
+            self.method, self.content_type, self.req_len, self.req_stream, url
+        )
+        return url + full_string
+
+
+# ============================================================================
+class TestPostQueryExtract:
+    @classmethod
+    def setup_class(cls):
+        cls.post_data = b"foo=bar&dir=%2Fbaz"
+        cls.binary_post_data = (
+            b"\x816l`L\xa04P\x0e\xe0r\x02\xb5\x89\x19\x00fP\xdb\x0e\xb0\x02,"
+        )
+
+    def test_post_extract_1(self):
+        mq = MethodQueryCanonicalizer(
+            "POST",
+            "application/x-www-form-urlencoded",
+            len(self.post_data),
+            BytesIO(self.post_data),
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST&foo=bar&dir=/baz"
+        )
+
+        assert (
+            mq.append_query("http://example.com/?123=ABC")
+            == "http://example.com/?123=ABC&__wb_method=POST&foo=bar&dir=/baz"
+        )
+
+    def test_post_extract_json(self):
+        post_data = b'{"a": "b", "c": {"a": 2}, "d": "e"}'
+        mq = MethodQueryCanonicalizer(
+            "POST", "application/json", len(post_data), BytesIO(post_data)
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST&a=b&a.2_=2&d=e"
+        )
+
+    def test_post_extract_json_top_list(self):
+        post_data = (
+            b'[{"a": "b", "c": {"a": 2}}, {"d": "e"}, "ignored", false, null, 0]'
+        )
+        mq = MethodQueryCanonicalizer(
+            "POST", "application/json", len(post_data), BytesIO(post_data)
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST&a=b&a.2_=2&d=e"
+        )
+
+    def test_post_extract_json_lines(self):
+        post_data = b'{"a": "b"}\n{"c": {"a": 2}, "d": "e"}'
+        mq = MethodQueryCanonicalizer(
+            "POST", "application/json", len(post_data), BytesIO(post_data)
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST&a=b&a.2_=2&d=e"
+        )
+
+    def test_put_extract_method(self):
+        mq = MethodQueryCanonicalizer(
+            "PUT",
+            "application/x-www-form-urlencoded",
+            len(self.post_data),
+            BytesIO(self.post_data),
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=PUT&foo=bar&dir=/baz"
+        )
+
+    def test_post_extract_non_form_data_1(self):
+        mq = MethodQueryCanonicalizer(
+            "POST",
+            "application/octet-stream",
+            len(self.post_data),
+            BytesIO(self.post_data),
+        )
+
+        # base64 encoded data
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST&__wb_post_data=Zm9vPWJhciZkaXI9JTJGYmF6"
+        )
+
+    def test_post_extract_non_form_data_2(self):
+        mq = MethodQueryCanonicalizer(
+            "POST", "text/plain", len(self.post_data), BytesIO(self.post_data)
+        )
+
+        # base64 encoded data
+        assert (
+            mq.append_query("http://example.com/pathbar?id=123")
+            == "http://example.com/pathbar?id=123&__wb_method=POST&__wb_post_data=Zm9vPWJhciZkaXI9JTJGYmF6"
+        )
+
+    def test_post_extract_length_invalid_ignore(self):
+        mq = MethodQueryCanonicalizer(
+            "POST", "application/x-www-form-urlencoded", 0, BytesIO(self.post_data)
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST"
+        )
+
+        mq = MethodQueryCanonicalizer(
+            "POST", "application/x-www-form-urlencoded", "abc", BytesIO(self.post_data)
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST"
+        )
+
+    def test_post_extract_length_too_short(self):
+        mq = MethodQueryCanonicalizer(
+            "POST",
+            "application/x-www-form-urlencoded",
+            len(self.post_data) - 4,
+            BytesIO(self.post_data),
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST&foo=bar&dir=%2"
+        )
+
+    def test_post_extract_length_too_long(self):
+        mq = MethodQueryCanonicalizer(
+            "POST",
+            "application/x-www-form-urlencoded",
+            len(self.post_data) + 4,
+            BytesIO(self.post_data),
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST&foo=bar&dir=/baz"
+        )
+
+    def test_post_extract_malformed_form_data(self):
+        mq = MethodQueryCanonicalizer(
+            "POST",
+            "application/x-www-form-urlencoded",
+            len(self.binary_post_data),
+            BytesIO(self.binary_post_data),
+        )
+
+        # base64 encoded data
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST&__wb_post_data=gTZsYEygNFAO4HICtYkZAGZQ2w6wAiw="
+        )
+
+    def test_post_extract_no_boundary_in_multipart_form_mimetype(self):
+        mq = MethodQueryCanonicalizer(
+            "POST", "multipart/form-data", len(self.post_data), BytesIO(self.post_data)
+        )
+
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=POST&__wb_post_data=Zm9vPWJhciZkaXI9JTJGYmF6"
+        )
+
+    def test_options(self):
+        mq = MethodQueryCanonicalizer("OPTIONS", "", 0, BytesIO())
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=OPTIONS"
+        )
+
+    def test_head(self):
+        mq = MethodQueryCanonicalizer("HEAD", "", 0, BytesIO())
+        assert (
+            mq.append_query("http://example.com/")
+            == "http://example.com/?__wb_method=HEAD"
+        )