Fix media type inference for URLs with query parameters

[fedexman]

Fix media type inference for URLs with query parameters

When using presigned URLs (eg AWS S3) with ImageUrl, AudioUrl, or VideoUrl, the media type inference fails

https://pics.s3.ap-northeast-1.amazonaws.com/test/Capture-2025-11-21-112402.png?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=fdafdas%2F20251121%2Fap-northeast-1%2Fs3%2Faws4_request&X-Amz-Date=20251121T023200Z&X-Amz-Expires=3600&X-Amz-SignedHeaders=host&X-Amz-Security-Token=IQoJb3JpZ2luX2VjEDoaD

to the agent we get this error,

Internal server error: Could not infer media type from image URL: ... Explicitly provide a `media_type` instead

the reason is that the _infer_media_type function only check the end of the url with url.endswith('.mkv') but do not parse the url.

I propose to parse the url with

from urllib.parse import urlparse path = urlparse(self.url).path if path.endswith('.mkv'): return 'video/x-matroska'

[Viicos]

Why aren't we using the mimetypes stdlib module? mimetypes.guess_type() already parses URLs and the current implementation doesn't take into account case insensitivity, etc.

[DouweM]

@Viicos Interestingly we already use that in DocumentUrl._infer_media_type, after checking a bunch of types ourselves :/

@fedexman Can you see if we can use mimetypes.guess_type() for all of these?

The method can be changed to just return str rather than XMediaType, as I don't think that type is used on any public fields.

[DouweM]

@Viicos Interestingly we already use that in DocumentUrl._infer_media_type, after checking a bunch of types ourselves :/

@fedexman Can you see if we can use mimetypes.guess_type() for all of these?

The method can be changed to just return str rather than XMediaType, as I don't think that type is used on any public fields.

[github-actions]

This PR is stale, and will be closed in 3 days if no reply is received.

[fedexman]

I'll update this week 🙏🙏

[fedexman]

I refactored using mimetypes. Some types are defined in the standard library, if not, they use some os files that is dependent of the machine. For all the types not in the standard library I added them manually to have reliable behavior.
Good for rereview 🙇

[DouweM]

@fedexman Please have a look at @Viicos's comments!

[fedexman]

Sorry for slow updates, I've used Mimetypes library in the past and it hides a lot of unexpected behaviors.
For example I have a test failing on pydantic ai due to this behavior:

import mimetypes filename = "data.xml" global_guess, _ = mimetypes.guess_type(filename) db = mimetypes.MimeTypes() db_guess, _ = db.guess_type(filename) print(f"Global: {global_guess}") print(f"Fresh: {db_guess}") assert global_guess == db_guess

Output

Global: application/xml Fresh: text/xml Traceback (most recent call last): File "<python-input-2>", line 13, in <module> assert global_guess == db_guess 

Even though the official docs says

class mimetypes.MimeTypes(filenames=(), strict=True)
This class represents a MIME-types database. By default, it provides access to the same database as the rest of this module. The initial database is a copy of that provided by the module, and may be extended by loading additional mime.types-style files into the database using the read() or readfp() methods. The mapping dictionaries may also be cleared before loading additional data if the default data is not desired.
The optional filenames parameter can be used to cause additional files to be loaded “on top” of the default database.

I get this test failing even though the xml test was guessed by this before

tests/test_messages.py:689 test_binary_content_from_path │ - assert binary_content == snapshot(BinaryContent(data=b'<think>about trains</think>', │ │ media_type='application/xml')) │ │ + assert binary_content == snapshot(BinaryContent(data=b'<think>about trains</think>', media_type='text/xml')) │ 

The mimetypes library has a internal _db with newer types, but using a fresh db doesn't contains those and use the os definitions.

I added the new application/xml manually but doing like we might risk using also other olds os definitions. Currently looking for better solution, i cannot find a clean way of creating a new mimetype db with the exact same state as the mimetype internal db

[fedexman]

Now tests should be passing and I added Vico's suggestions. Tell me if you have an idea of cleaner solution for the mime type db init

[DouweM]

@fedexman Can you have a look at the failing tests please?

[Viicos]

Even though the official docs says

Yeah, this is annoying and a documentation update is pending at python/cpython#27750.

One way to do it is to just replicate how the default db is instantiated:

_mime_types = MimeTypes() # The default db is actually different from the ones we can manually instantiate # (see https://github.com/python/cpython/pull/27750). As such, replicate # what is being done in `mimetypes.init()`: _mime_types.read_windows_registry() for file in mimetypes.knownfiles: if os.path.isfile(file): _mime_types.read(file)

[Viicos]

These are not required now, right?

[fedexman]

I only added the one not defined natively by python or the one that have a different definition in python

[fedexman]

https://github.com/python/cpython/blob/a183a11db8bc2520c52814635de2df118d2d7e8c/Lib/mimetypes.py#L434C1-L604C1

[Viicos]

I'm not sure I understand: taking .mp4 as an example, it is defined in the mapping and maps to video/mp4?

[fedexman]

yes, but test are failing on 3.10 even though we add video/mp4 explicitly and it is also added in the python lib 😅 passing locally though so difficult to test. I'll update with test passing

[fedexman]

ok the query string ignore was supported only from 3.11 😅
python/cpython#66543
on 3.10 it is not supported

in 3.10 this code gives Result: (None, None)

import mimetypes import sys url = "https://example.com/image.png?token=123" result = mimetypes.guess_type(url) print(f"Result: {result}") 

[fedexman]

Sorry for not updating the PR. In case we are using a python feature that is not supported in 3.10 what is the common way to do, does it mean we cannot use mimetype for this task, or do we add a new branch logic for 3.10 😅 what is the usual solution for these cases ?
the purpose of this PR was to make things cleaner and support query types by using a native python feature, but if we cannot use it, this PR does not make so much sense. I can switch to a urllib.parse logic though 🤔 @Viicos

[Viicos]

Sorry should have come back here. Given that 3.10 is the oldest supported version and we will drop support in less than a year, I'd rather skip the test in this version. Prior to this PR, it wouldn't be supported in any Python version anyway, so I think this is fine.

[Viicos]

With that in mind, are the manually added mime types still necessary?

[fedexman]

yes i realized during debugging i may have added unnecessary types, i remade a check with https://github.com/python/cpython/blob/3.10/Lib/mimetypes.py and now it should not have any duplicates anymore with native python

[fedexman]

good for rereview @Viicos

[Viicos]

Thanks @fedexman 🙏