Roblox Cache Buster: Simple script to convert Roblox cache files in `%appdata%\Local\Roblox\rbx-storage` folder to normal, openable files (if valid). Run the script inside `%appdata%\Local\Roblox\rbx-storage` normally in Windows OR ~/.var/app/org.vinegarhq.Sober/cache/sober/http for Sober under Linux. Now tested in Linux and original Windows wit…

rb-cachebuster.py

from genericpath import isfile
import puremagic, io, os

os.makedirs('cachebuster', exist_ok=True)

# Tries to identify rbx file not identifiable by puremagic:
def rbx_format(data):
    head = data.read(20)
    if head.startswith(b'version'):
        # https://devforum.roblox.com/t/roblox-filemesh-format-specification/326114
        return 'mesh'
    elif head.startswith(b'\xabKTX 11'):
        # b'\xabKTX is a texture, https://registry.khronos.org/KTX/specs/1.0/ktxspec.v1.html
        return 'ktx'
    elif head.startswith(b'<roblox xmlns'):
        return 'xml'
    # b'<roblox!  is an animation?
    # roblox-assettypeid files: https://create.roblox.com/docs/reference/engine/enums/AssetType
    return

files = []
for file_name in os.listdir('.'):
    if not os.path.isfile(file_name):
        if len(file_name) == 2:
            files += [os.path.join(file_name, f) for f in os.listdir(file_name)]
    else:
        files.append(file_name)

for file_name in files:
    print(file_name)
    if not os.path.isfile(file_name):
        continue
    new_file_headers = b""
    new_file_content = None

    with open(file_name, mode='rb') as file:
        if file.read(4) != b"RBXH":
            continue

        print()
        # Seeking to HTTP request
        file.read(4)
        d = int.from_bytes(file.read(4), byteorder='little')
        print(d)
        url = file.read(d)
        print(url)

        sep = file.read(1)
        assert sep == b'\x00'
        file.read(4)  # not sure what this is?
        header_length = int.from_bytes(file.read(4), 'little')
        unk = file.read(12)
        #print("Header length:", header_length)
        # READ 03 00 00 00
        mark = int.from_bytes(file.read(4), 'little')
        assert mark == 3
        if header_length:
            header = file.read(header_length)
            print("Header:", header)

        # print("READ CONTENT")
        new_file_content = io.BytesIO(file.read())
        
        # print(puremagic.magic_stream(file))

    # print(new_file.read(16))
    new_file_name = None
    file_name = os.path.basename(file_name)
    puremagic_result = None
    try:
        puremagic_result = puremagic.magic_stream(new_file_content)
        # print(puremagic_result[0].extension)
        new_file_name = os.path.join("cachebuster", file_name + puremagic_result[0].extension)
        ftype = puremagic_result[0].name
    except:
        new_file_name = os.path.join("cachebuster", file_name)

    new_file_content.seek(0)

    if not new_file_content.read(1):
        print(file_name + ": No content")
        continue

    if not puremagic_result:
        new_file_content.seek(0)
        rbx_result = rbx_format(new_file_content)
        if rbx_result:
            new_file_name = os.path.join("cachebuster", file_name + "." + rbx_result)
            ftype = rbx_result
        else:
            print(file_name + ": OK, unknown")
            new_file_content.seek(0)
            print('FILE:', new_file_content.read(22))
            continue
        
    new_file_content.seek(0)

    with open(new_file_name, 'wb') as new_file:
        new_file.write(new_file_content.read())

    print(file_name + ": OK, " + ftype)

requirements.txt

puremagic==1.14

I originally forked this script to run in Linux (with Sober) and the file handling was not cross platform, My bad. Now it should work under Windows and Linux and in a directory that has cache files stored in the root dir, or in 00/ - ff/ sub-directories.

@SpacePiracy112 does this work for you now? Thx for your comment on https://gist.github.com/Hans5958/f9870ae89f80b5d972d95031e24584bb?permalink_comment_id=5694939#gistcomment-5694939

@hornc thank you, I would like to note that this script has significantly less occurrences where it exits upon encountering a problematic file type, compared to the original (while that one still worked)