handoff fixes

backend/app/services/mime_body_extractor.py

@@ -0,0 +1,106 @@
+from dataclasses import dataclass, field
+from email import policy
+from email.message import EmailMessage, Message
+from email.parser import BytesParser
+
+
+@dataclass
+class ExtractedMimePart:
+    filename: str
+    content_type: str
+    payload: bytes
+    content_id: str | None = None
+    disposition: str = "attachment"
+
+
+@dataclass
+class MirroredMimeContent:
+    body_text: str = ""
+    body_html: str | None = None
+    inline_parts: list[ExtractedMimePart] = field(default_factory=list)
+    file_attachments: list[ExtractedMimePart] = field(default_factory=list)
+
+
+def _part_payload(part: Message) -> bytes:
+    payload = part.get_payload(decode=True)
+    if isinstance(payload, bytes):
+        return payload
+    if isinstance(payload, str):
+        return payload.encode("utf-8", errors="replace")
+    return b""
+
+
+def _part_text(part: Message) -> str:
+    try:
+        value = part.get_content()
+    except (LookupError, UnicodeError):
+        value = _part_payload(part).decode("utf-8", errors="replace")
+    return value if isinstance(value, str) else ""
+
+
+def _content_id(part: Message) -> str | None:
+    header = part.get("Content-ID")
+    if not header:
+        return None
+    return header.strip().strip("<>")
+
+
+def _is_attachment_part(part: Message) -> bool:
+    disposition = (part.get_content_disposition() or "").lower()
+    if disposition == "attachment":
+        return True
+    if disposition == "inline":
+        return False
+    filename = part.get_filename()
+    content_type = part.get_content_type()
+    if filename and content_type not in {"text/plain", "text/html"}:
+        return True
+    return False
+
+
+def _walk_parts(message: Message, collector: MirroredMimeContent) -> None:
+    if message.is_multipart():
+        for part in message.get_payload():
+            if isinstance(part, Message):
+                _walk_parts(part, collector)
+        return
+
+    content_type = message.get_content_type()
+    cid = _content_id(message)
+    filename = message.get_filename() or (f"inline-{cid}" if cid else "attachment")
+
+    if content_type == "text/plain" and not _is_attachment_part(message):
+        text = _part_text(message).strip()
+        if text and not collector.body_text:
+            collector.body_text = text
+        return
+
+    if content_type == "text/html" and not _is_attachment_part(message):
+        html = _part_text(message).strip()
+        if html:
+            collector.body_html = html
+        return
+
+    payload = _part_payload(message)
+    if not payload:
+        return
+
+    extracted = ExtractedMimePart(
+        filename=filename,
+        content_type=content_type,
+        payload=payload,
+        content_id=cid,
+        disposition="inline" if cid or message.get_content_disposition() == "inline" else "attachment",
+    )
+
+    if cid or extracted.disposition == "inline":
+        collector.inline_parts.append(extracted)
+    elif _is_attachment_part(message):
+        collector.file_attachments.append(extracted)
+
+
+def extract_mirror_content(raw: bytes) -> MirroredMimeContent:
+    message = BytesParser(policy=policy.default).parsebytes(raw)
+    collector = MirroredMimeContent()
+    _walk_parts(message, collector)
+    return collector