feat(discord): allow_any_attachment config to accept arbitrary file types

The Discord adapter silently dropped any attachment whose extension wasn't
in the SUPPORTED_DOCUMENT_TYPES allowlist (PDF, text family, zip, office).
Users uploading .wav / .bin / other unrecognized formats saw nothing in
their conversation — the file got logged as 'Unsupported document type'
and discarded before the agent ever saw it.

Add discord.allow_any_attachment (default false) to bypass the allowlist.
When on:
  - Any file is downloaded, cached under ~/.hermes/cache/documents/, and
    surfaced as a DOCUMENT-typed event with application/octet-stream MIME
  - gateway/run.py already emits a context note with the cached path,
    auto-translated via to_agent_visible_cache_path() for Docker/Modal
    sandboxed terminals
  - File body is NOT inlined — only the path — so binary uploads don't
    blow up the context window
  - Allowlisted text formats (.txt/.md/.log) keep their 100 KiB inline
    behavior unchanged

Also adds discord.max_attachment_bytes (default 32 MiB matches the
historical hardcoded cap; 0 = unlimited) since users opting into arbitrary
types may want to raise the cap. The whole attachment is held in memory
while being cached, so unlimited carries a real memory cost.

Env overrides: DISCORD_ALLOW_ANY_ATTACHMENT, DISCORD_MAX_ATTACHMENT_BYTES.

Discord-only by deliberate scope. Telegram has hard 20 MB API limits and
Slack has its own caps — extending the same flag there is a separate
follow-up if/when requested.
This commit is contained in:
teknium1 2026-05-16 20:25:47 -07:00 committed by Teknium
parent 3b39096904
commit 407a11b419
5 changed files with 258 additions and 10 deletions

View file

@ -384,3 +384,148 @@ class TestIncomingDocumentHandling:
assert event.message_type == MessageType.PHOTO
assert event.media_urls == ["/tmp/cached_image.png"]
assert event.media_types == ["image/png"]
class TestAllowAnyAttachment:
"""Cover the discord.allow_any_attachment config flag.
With the flag off (default), unknown file types are dropped. With it on,
they get cached and surfaced to the agent as DOCUMENT events with
application/octet-stream MIME so gateway/run.py emits a path-pointing
context note.
"""
@pytest.mark.asyncio
async def test_unknown_type_skipped_by_default(self, adapter):
"""Default (flag off): unknown extension is dropped.
With no text + no cached media, the adapter may legitimately decline
to dispatch the event at all, so we don't assert on call_args here —
we just verify the file wasn't cached.
"""
with _mock_aiohttp_download(b"should not be cached"):
msg = make_message([
make_attachment(filename="weird.xyz", content_type="application/x-custom")
])
await adapter._handle_message(msg)
if adapter.handle_message.call_args is not None:
event = adapter.handle_message.call_args[0][0]
assert event.media_urls == []
@pytest.mark.asyncio
async def test_unknown_type_cached_when_flag_on(self, adapter):
"""Flag on: unknown extension is cached as application/octet-stream."""
adapter.config.extra["allow_any_attachment"] = True
with _mock_aiohttp_download(b"\x00\x01\x02 binary payload"):
msg = make_message([
make_attachment(filename="weird.xyz", content_type="application/x-custom")
])
await adapter._handle_message(msg)
event = adapter.handle_message.call_args[0][0]
assert len(event.media_urls) == 1
assert os.path.exists(event.media_urls[0])
# Falls back to the source content_type when we have one.
assert event.media_types == ["application/x-custom"]
assert event.message_type == MessageType.DOCUMENT
# We deliberately do NOT inline arbitrary bytes — run.py emits the
# path-pointing note based on DOCUMENT + octet-stream MIME.
assert "[Content of" not in (event.text or "")
@pytest.mark.asyncio
async def test_unknown_type_no_content_type_becomes_octet_stream(self, adapter):
"""Flag on + no content_type from discord: MIME falls back to octet-stream."""
adapter.config.extra["allow_any_attachment"] = True
with _mock_aiohttp_download(b"raw bytes"):
msg = make_message([
make_attachment(filename="mystery.bin", content_type=None)
])
await adapter._handle_message(msg)
event = adapter.handle_message.call_args[0][0]
assert event.message_type == MessageType.DOCUMENT
assert event.media_types == ["application/octet-stream"]
@pytest.mark.asyncio
async def test_max_attachment_bytes_caps_uploads(self, adapter):
"""discord.max_attachment_bytes overrides the historical 32 MiB cap."""
adapter.config.extra["allow_any_attachment"] = True
adapter.config.extra["max_attachment_bytes"] = 1024 # 1 KiB
msg = make_message([
make_attachment(
filename="too_big.xyz",
content_type="application/x-custom",
size=2048,
)
])
await adapter._handle_message(msg)
event = adapter.handle_message.call_args[0][0]
assert event.media_urls == []
@pytest.mark.asyncio
async def test_max_attachment_bytes_zero_means_unlimited(self, adapter):
"""max_attachment_bytes=0 disables the size cap entirely."""
adapter.config.extra["allow_any_attachment"] = True
adapter.config.extra["max_attachment_bytes"] = 0
# 64 MiB — would normally exceed the historical 32 MiB hardcoded cap.
with _mock_aiohttp_download(b"x" * 16):
msg = make_message([
make_attachment(
filename="huge.xyz",
content_type="application/x-custom",
size=64 * 1024 * 1024,
)
])
await adapter._handle_message(msg)
event = adapter.handle_message.call_args[0][0]
assert len(event.media_urls) == 1
@pytest.mark.asyncio
async def test_allowlisted_doc_unchanged_when_flag_on(self, adapter):
"""Flag on must not change handling of types already in SUPPORTED_DOCUMENT_TYPES.
A .txt should still get its content inlined (the historical behavior),
and the MIME should still be the canonical text/plain not whatever
discord guessed.
"""
adapter.config.extra["allow_any_attachment"] = True
file_content = b"still a text file"
with _mock_aiohttp_download(file_content):
msg = make_message(
attachments=[make_attachment(filename="notes.txt", content_type="text/plain")],
content="check this",
)
await adapter._handle_message(msg)
event = adapter.handle_message.call_args[0][0]
assert "[Content of notes.txt]:" in event.text
assert "still a text file" in event.text
assert event.media_types == ["text/plain"]
def test_helper_reads_env_fallback(self, adapter, monkeypatch):
"""Helper falls back to DISCORD_ALLOW_ANY_ATTACHMENT env var."""
assert adapter._discord_allow_any_attachment() is False
monkeypatch.setenv("DISCORD_ALLOW_ANY_ATTACHMENT", "true")
assert adapter._discord_allow_any_attachment() is True
monkeypatch.setenv("DISCORD_ALLOW_ANY_ATTACHMENT", "no")
assert adapter._discord_allow_any_attachment() is False
def test_helper_config_overrides_env(self, adapter, monkeypatch):
"""config.yaml setting wins over env var."""
monkeypatch.setenv("DISCORD_ALLOW_ANY_ATTACHMENT", "true")
adapter.config.extra["allow_any_attachment"] = False
assert adapter._discord_allow_any_attachment() is False
def test_max_bytes_helper_invalid_value_falls_back(self, adapter):
"""Garbage in max_attachment_bytes config falls back to 32 MiB."""
adapter.config.extra["max_attachment_bytes"] = "not-a-number"
assert adapter._discord_max_attachment_bytes() == 32 * 1024 * 1024