Skip to content

Your lightweight open-source personal agent.

fix(media): keep upload filenames within the UTF-8 byte limit

Shoaibb4f381ca5581
diff --git a/nanobot/utils/media_decode.py b/nanobot/utils/media_decode.py
index cc8494bbe..5d39226f4 100644
--- a/nanobot/utils/media_decode.py
+++ b/nanobot/utils/media_decode.py
@@ -92,5 +92,8 @@ def media_destination(media_dir: Path, mime_type: str, *, filename: str | None =
ext = _MIME_EXTENSION_OVERRIDES.get(mime_type) or mimetypes.guess_extension(mime_type) or ".bin"
base = safe_filename(filename or "")
stem = Path(base).stem[:80] if base else ""
+ # Leave room for the 12-character ID, separator, and extension.
+ max_stem_bytes = 255 - 13 - len(ext.encode("utf-8"))
+ stem = stem.encode("utf-8")[:max_stem_bytes].decode("utf-8", errors="ignore")
saved_name = f"{uuid.uuid4().hex[:12]}_{stem}{ext}" if stem else f"{uuid.uuid4().hex[:12]}{ext}"
return media_dir / safe_filename(saved_name)
diff --git a/tests/utils/test_media_decode.py b/tests/utils/test_media_decode.py
index cbe6cd187..3df6bfa5c 100644
--- a/tests/utils/test_media_decode.py
+++ b/tests/utils/test_media_decode.py
@@ -62,6 +62,24 @@ def test_saves_document_with_safe_original_name(tmp_path) -> None:
assert Path(result).read_bytes() == b"%PDF-1.4"
[email protected]("stem", ["报告" * 40, "📄" * 60], ids=["cjk", "emoji"])
+def test_saves_document_with_long_unicode_name(tmp_path: Path, stem: str) -> None:
+ payload = b"%PDF-1.4"
+ result = save_base64_data_url(
+ _data_url(payload, mime="application/pdf"),
+ tmp_path,
+ filename=f"{stem}.pdf",
+ )
+
+ assert result is not None
+ saved = Path(result)
+ assert saved.parent == tmp_path
+ assert saved.suffix == ".pdf"
+ assert len(saved.name.encode("utf-8")) <= 255
+ assert "�" not in saved.name
+ assert saved.read_bytes() == payload
+
+
def test_document_filename_cannot_escape_media_dir(tmp_path) -> None:
result = save_base64_data_url(
_data_url(b"%PDF-1.4", mime="application/pdf"),