Self-improving AI Employees for everyone
Python ai-agentsai-employeeai-employeesassistantautonomous-agentsco-employeescompanygrok-botinstinctloop-agentsmuseown-your-datapersonalself-hosted
"""Turn a discord.py message object into a flat event payload.
Every snowflake is stringified here. discord.py exposes them as int, while
Discord's own JSON uses strings, and an id past 2^53 loses precision the
moment anything treats it as a number -- edge conditions comparing an id are
the most natural thing a user writes, and two distinct 18-digit ids comparing
equal is a bug with no visible cause.
Attachments travel as metadata, never bytes. A node result is persisted,
broadcast, and replayed into LLM context, so the payload is a reference and
discordAction fetches it on demand.
"""
from __future__ import annotations
from typing import Any, Dict, List, Optional
from core.logging import get_logger
logger = get_logger(__name__)
# Enough to fetch the file later and to decide whether it is worth fetching.
# Deliberately scalar-only: no nested media objects, no thumbnails.
_ATTACHMENT_FIELDS = ("id", "filename", "size", "url", "content_type", "width", "height")
def _sid(value: Any) -> Optional[str]:
"""Snowflake to string, preserving None."""
return None if value is None else str(value)
def shape_attachment(attachment: Any) -> Dict[str, Any]:
payload: Dict[str, Any] = {}
for field in _ATTACHMENT_FIELDS:
value = getattr(attachment, field, None)
if value is not None:
payload[field] = _sid(value) if field == "id" else value
return payload
def shape_message(message: Any, *, account_id: str) -> Dict[str, Any]:
"""Flatten one message into the trigger's output shape."""
author = getattr(message, "author", None)
guild = getattr(message, "guild", None)
channel = getattr(message, "channel", None)
reference = getattr(message, "reference", None)
attachments = [shape_attachment(a) for a in (getattr(message, "attachments", None) or [])]
return {
"account_id": account_id,
"message_id": _sid(getattr(message, "id", None)),
"channel_id": _sid(getattr(channel, "id", None)),
"channel_name": getattr(channel, "name", None),
"guild_id": _sid(getattr(guild, "id", None)),
"guild_name": getattr(guild, "name", None),
# None for a DM, which is also how a consumer tells the two apart.
"is_dm": guild is None,
"author_id": _sid(getattr(author, "id", None)),
"author_name": getattr(author, "name", None),
"author_display_name": getattr(author, "display_name", None),
"author_is_bot": bool(getattr(author, "bot", False)),
"content": getattr(message, "content", "") or "",
"timestamp": (
message.created_at.isoformat() if getattr(message, "created_at", None) else None
),
"attachments": attachments,
"has_attachments": bool(attachments),
"mentions_me": _mentions_me(message),
"reply_to_message_id": _sid(getattr(reference, "message_id", None)),
}
def _mentions_me(message: Any) -> bool:
"""Whether the bot was mentioned.
Useful as a filter because without the message-content intent this is one
of the few signals that still arrives populated.
"""
try:
state = getattr(message, "_state", None)
me = getattr(state, "user", None) if state else None
if me is None:
return False
return any(getattr(u, "id", None) == me.id for u in (message.mentions or []))
except Exception:
return False
async def dispatch_message(message: Any, *, account_id: str) -> None:
"""Shape and emit one inbound message."""
from ._events import dispatch_discord_message_received
payload = shape_message(message, account_id=account_id)
await dispatch_discord_message_received(payload)
__all__ = ["dispatch_message", "shape_attachment", "shape_message"]