fix(gateway): allow [[as_document]] glued directly to MEDIA path

[liuhao1024]

What does this PR do?

Fixes a regex bug in MEDIA_TAG_CLEANUP_RE that caused MEDIA tags to be silently ignored when a directive like [[as_document]] was concatenated directly to the file extension without whitespace.

When the regex failed to match, the MEDIA tag was not stripped and the file path was not extracted, resulting in the file not being delivered while the gateway reported success (no error surfaced).

Root cause

The lookahead character class [\s"',;:)}]|$)inMEDIA_TAG_CLEANUP_REdid not include[. When [[as_document]]immediately followed the file extension (e.g.,MEDIA:/home/user/report.xlsx[[as_document]]), the lookahead assertion failed because [` was not a permitted boundary character.

Fix

Add \[ to the lookahead character class so it becomes [\s"',;:)}[]|$)`. This allows directives to follow the path without whitespace.

This change is safe because [ characters (specifically [[as_document]] and [[as_image]]) are already stripped elsewhere in the same file via a literal .replace("[[as_document]]", "") call.

Related Issue

Fixes #63632

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)

Changes Made

• gateway/platforms/base.py: Added \[ to the lookahead character class in MEDIA_TAG_CLEANUP_RE
• tests/gateway/test_media_tag_cleanup.py: New test file with 3 regression tests

How to Test

Run the new regression tests:

pytest tests/gateway/test_media_tag_cleanup.py -v

Observed result: All 3 tests pass, confirming that:

1. MEDIA:/home/hermes/report.xlsx[[as_document]] now matches (fixes the issue)
2. MEDIA:/home/hermes/report.xlsx [[as_document]] still matches (baseline unchanged)
3. MEDIA:/tmp/file.docx at end of string still matches (baseline unchanged)

You can also verify the fix with a quick Python REPL session:

from gateway.platforms.base import MEDIA_TAG_CLEANUP_RE

# Issue case: [[as_document]] glued directly
text = "Готово. MEDIA:/home/hermes/report.xlsx[[as_document]]"
assert MEDIA_TAG_CLEANUP_RE.search(text) is not None  # Now passes
stripped = MEDIA_TAG_CLEANUP_RE.sub("", text)
assert "MEDIA:" not in stripped  # Tag is removed
assert "/home/hermes/report.xlsx" not in stripped  # Path is removed

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/gateway/test_media_tag_cleanup.py -v and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 15.2

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

[teknium1]

Thanks for isolating a real media-extraction edge case. The premise still holds on current main: extract_media() scans the original response at gateway/platforms/base.py:3684-3687, while MEDIA_TAG_CLEANUP_RE at gateway/platforms/base.py:1503 does not currently accept [ after the extension.

Problems

• The proposed [ boundary at gateway/platforms/base.py:1495 is broader than the supported directive grammar. It also accepts arbitrary bracket-bearing filenames, allowing the regex to match a valid-extension prefix rather than leaving an unrecognized path intact.
• tests/gateway/test_media_tag_cleanup.py tests only MEDIA_TAG_CLEANUP_RE. The delivery path is BasePlatformAdapter.extract_media() (gateway/platforms/base.py:3637-3731), and current code supports [[as_document]], not [[as_image]] (gateway/platforms/base.py:1542, 3667-3671).

Suggested changes

• Match the literal [[as_document]] boundary rather than a generic [, and add a production-path regression test in tests/gateway/test_platform_base.py asserting both extraction and cleaned output.

Automated hermes-sweeper review.

[teknium1]

Please restrict this new boundary to the literal supported [[as_document]] directive rather than any [. A generic bracket boundary can make MEDIA:/safe/report.xlsx[revision] match and deliver the .xlsx prefix, whereas the current anchored parser leaves that unrecognized path intact.

[teknium1]

[[as_image]] is not a supported directive on current main: cleanup only strips [[audio_as_voice]] and [[as_document]]. Please remove this case or implement that directive separately; add an extract_media() regression for the supported glued [[as_document]] form instead.

[teknium1]

Merged via PR #72170 (rebase-merge) — your commit was cherry-picked onto current main with your authorship preserved in git history. Thanks for the fix!

PR #72170 consolidated six open MEDIA-delivery fixes (formatting-variant regex misses + the Discord upload race) into one salvage wave so the overlapping regex changes could be resolved together.