""" Chatto Platform Adapter for Hermes Agent. A plugin-based gateway adapter that connects to a Chatto server (self-hosted team chat) and relays messages to/from the Hermes agent. The adapter uses the chattolib library for all Chatto API interactions, including both outbound messaging and realtime WebSocket connections. """ from __future__ import annotations import random from gateway.platforms.helpers import MessageDeduplicator # Put the vendored dependencies for THIS platform on sys.path before importing # anything from chattolib. Imported relatively as part of the plugin package and # absolutely when this module is loaded standalone (e.g. by the tests). try: from .vendor_path import setup_vendor_path except ImportError: # pragma: no cover - depends on how the module is loaded from vendor_path import setup_vendor_path setup_vendor_path() import asyncio import hashlib import logging import mimetypes import os from datetime import datetime, timezone from enum import StrEnum from typing import Any, Dict, List, Literal, Optional, Tuple, cast from urllib.parse import urlsplit logger = logging.getLogger(__name__) from gateway.platforms.base import ( BasePlatformAdapter, SendResult, MessageEvent, MessageType, ProcessingOutcome, cache_media_bytes, get_inbound_media_max_bytes, validate_inbound_media_size, ) from gateway.config import Platform, PlatformConfig # Chattolib imports (vendored) # Using vendored chattolib from vendor/chattolib/ # See vendor_chattolib.sh for how to update the vendored copy # Absolute imports — vendor/ is on sys.path (see above) and chattolib's own # modules import each other absolutely. Mixing in relative ".vendor.chattolib" # imports would load a second, distinct copy of every module, so isinstance() # checks across the two copies would silently fail. try: from chattolib.client import ( ChattoClient, ) from chattolib.exceptions import ( ChattoAuthError, ChattoError, ) from chattolib.realtime import ( ChattoRealtimeError, ChattoRealtimeCloseError, RealtimeEvent, stream_events ) from chattolib.realtime_types import ( MessagePostedPayload, ReactionPayload, ) from chattolib.types import ( PresenceStatus, RoomKind, User ) except ImportError as e: # Fail loudly: continuing here only defers the failure to a confusing # NameError somewhere deep in the adapter. logger.error("Chatto: failed to import vendored chattolib: %s", e) raise try: from .platform_config import ( ChattoConfiguration, ChattoConstants, ) except ImportError: # pragma: no cover - loaded as a top-level module (tests) from platform_config import ( ChattoConfiguration, ChattoConstants, ) # --------------------------------------------------------------------------- # # Chat types # --------------------------------------------------------------------------- # class HermesChatType(StrEnum): """The ``chat_type`` vocabulary the Hermes gateway understands. Declared in ``gateway/session.py:161`` as ``"dm", "group", "channel", "thread"`` and consumed as a bare string all over the gateway: ``SessionSource.description`` (session.py:239) and the PII-redacting description in ``build_session_context_prompt`` (session.py:537) both branch on these exact values and fall back to a nameless generic case for anything else, and ``build_session_key`` puts the value straight into the session key. Passing a chattolib ``RoomKind`` (``"ROOM_KIND_CHANNEL"``) therefore does not fail loudly — it just quietly degrades what the agent is told about where it is. A StrEnum so it stays a drop-in ``str`` at every one of those call sites. GROUP vs CHANNEL ---------------- There is no strict contract between the two, and the adapters disagree in practice: Slack labels every non-DM conversation ``"group"`` (including real channels), Discord uses both, and Telegram reserves ``"channel"`` for actual broadcast channels. The intended reading is ``group`` = ordinary multi-participant chat, ``channel`` = broadcast surface. The distinction only changes behaviour in three places: 1. Authorization (``gateway/authz_mixin.py``) — the only security-relevant one. The group-scoped env allowlists apply to ``{"group", "forum"}`` ONLY, never to ``"channel"``: ``{PLATFORM}_GROUP_ALLOWED_USERS`` / ``_GROUP_ALLOWED_CHATS`` (:616), the chat-id allowlist (:708) and the Telegram legacy shim (:724). The adapter-delegation paths in turn treat all three alike (:461, :649, :674, :694), where the value only picks ``group_allow_from`` over ``allow_from`` from ``config.extra``. For Chatto both choices are equivalent today: those group env maps hold Telegram and QQBot only (:535-541), and our own allowlist runs through ``CHATTO_ALLOWED_USERS``, which is chat_type-independent. 2. What the agent is told — ``SessionSource.description`` renders ``"group: Name"`` vs ``"channel: Name"`` (session.py:239-246), likewise the PII-redacted variant (session.py:537-544). 3. The session key, which embeds the literal (session.py:1192). Changing the value for a room re-buckets its existing sessions. Explicitly NOT affected: ``is_shared_multi_user_session`` (session.py:1063) only looks at ``"dm"`` and ``thread_id``, so sender prefixes, the multi-user prompt line and ``group_sessions_per_user`` treat group and channel identically. """ DM = "dm" GROUP = "group" CHANNEL = "channel" # Emitted by adapters whose thread events are their own chat type (Slack, # Discord). We don't: a Chatto thread keeps its room's chat_type and is # identified by ``thread_id`` on the source instead. Listed for the record, # because build_session_key rewrites the slot to "thread" itself # (session.py:1190). THREAD = "thread" # Not declared in session.py:161 but real: Telegram forum topics travel as # "forum", and the authz group allowlists above accept it alongside "group". # Chatto has no equivalent, so we never emit it. # Chatto only distinguishes DMs from channels. UNSPECIFIED means the server # sent a kind this vendored chattolib doesn't know: map it to the generic # multi-user bucket rather than guessing "channel", and never to "dm" — that # value drives session isolation (is_shared_multi_user_session, session.py:1063) # and would silently turn a room into a private conversation. # # CHANNEL for RoomKind.CHANNEL is the descriptive choice and carries no # behavioural cost (see the GROUP vs CHANNEL note above). Switching to GROUP for # Slack parity would be this one line — plus the re-bucketing of existing # sessions that point 3 of that note describes. _ROOM_KIND_TO_CHAT_TYPE: Dict[RoomKind, HermesChatType] = { RoomKind.DM: HermesChatType.DM, RoomKind.CHANNEL: HermesChatType.CHANNEL, RoomKind.UNSPECIFIED: HermesChatType.GROUP, } def chat_type_for_room_kind(kind: Optional[RoomKind]) -> HermesChatType: """Map a chattolib RoomKind onto the gateway's chat_type vocabulary. An unknown or missing kind becomes ``GROUP`` — see ``_ROOM_KIND_TO_CHAT_TYPE``. """ if kind is None: return HermesChatType.GROUP return _ROOM_KIND_TO_CHAT_TYPE.get(kind, HermesChatType.GROUP) # --------------------------------------------------------------------------- # # Adapter # --------------------------------------------------------------------------- # def hermes_adapter_factory(config: PlatformConfig): """Factory wrapper that constructs ChattoAdapter from a PlatformConfig.""" return ChattoAdapter(config) class ChattoAdapter(BasePlatformAdapter): """Chatto platform adapter — receives messages via WebSocket realtime, sends via ConnectRPC.""" _SPLIT_THRESHOLD = 9900 # Read by BasePlatformAdapter.max_message_length_for_chat(), which the # gateway and the stream consumer use to chunk outgoing messages. Without # it they fall back to 4096 and split Chatto messages far earlier than # necessary — send() itself already truncates at 10000. MAX_MESSAGE_LENGTH = ChattoConstants.MAX_MESSAGE_LENGTH splits_long_messages = True supports_code_blocks: bool = True supports_status_text: bool = True # client.update_custom_status def __init__(self, pconfig: PlatformConfig): """Signature needs to be compatible with BasePlatformAdapter.__init__ """ super().__init__(config=pconfig, platform=Platform(ChattoConstants.PLATFORM_NAME)) # "extra" has been pre-processed by Hermes-Framework to be a dict of extra config values insode PlatformConfig. # --- Configuration from our configuration data class with some logic --- self.chatto_config: ChattoConfiguration = ChattoConfiguration(pconfig) # ------ State ------- # SDK runtime handle (injected by Hermes); annotate for Pylance self.sdk: Any = getattr(self, "sdk", None) # Our own user, filled in by connect(). Events arriving before connect() # completes must not blow up on an undefined attribute. self.me: Optional[User] = None # --- Runtime state --- self._user_id: str = "" self._user_display: str = "" self._room_names: Dict[str, str] = {} self._room_kinds: Dict[str, RoomKind] = {} self._our_thread_roots: set = set() # thread root event IDs we created self._our_message_ids: set = set() # message IDs we sent (for thread root detection) self._seen: list[str] = [] # Plain RealtimeEvent-id list self._resume_cursor: Optional[str] = None self._watch_room_ids: List[str] = [] self._ws_task: Optional[asyncio.Task] = None self._presence_task: Optional[asyncio.Task] = None self._ws_ready: Optional[asyncio.Event] = None self._ws_active = False self._ws_ref = None # reference to open websocket for dynamic resubscribe # Persistent typing indicator loops per room self._typing_tasks: Dict[str, asyncio.Task] = {} # Member directory cache: user_id -> user info dict self._user_cache: Dict[str, User] = {} # Chattolib client cache and lock for async access. self._chatto_client: Optional[ChattoClient] = None self._chatto_client_lock: asyncio.Lock = asyncio.Lock() # Dedup — chattolib may redeliver events across reconnects. self._dedup = MessageDeduplicator() # ------------------------------------------------------------------ # # Auth # ------------------------------------------------------------------ # async def _get_chatto_client(self: ChattoAdapter) -> Optional[ChattoClient]: """Get or create a ChattoClient instance.""" if self._chatto_client is not None: return self._chatto_client async with self._chatto_client_lock: if self._chatto_client is not None: return self._chatto_client try: assert(self.chatto_config.login.value) # now we can assume, _login is available. client = await self._open_client( base_url=self.chatto_config.base_url.value, login=self.chatto_config.login.value, password=self.chatto_config.password.value, token=self.chatto_config.token.value, ) self._chatto_client = client self._token = client.token logger.info("Chatto: logged in as '%s' via chattolib", self.chatto_config.login.value) return client except ChattoAuthError as e: logger.error("Chatto: authentication failed: %s", e) return None except (ChattoError, ValueError) as e: logger.error("Chatto: failed to create client: %s", e) return None async def _require_client(self) -> ChattoClient: """Return a ChattoClient or raise RuntimeError if unavailable. Use this helper when the caller expects a client to exist and wants a single canonical failure path. Methods that prefer a soft-fail can catch RuntimeError and return gracefully. """ client = await self._get_chatto_client() if client is None: raise RuntimeError("Chatto client unavailable") return client async def _ensure_token(self) -> bool: """Ensure we have a logged-in Chatto client and token.""" if self.chatto_config.token.value and isinstance(self._chatto_client, ChattoClient): return True client = await self._get_chatto_client() return client is not None # ------------------------------------------------------------------ # # Connection # ------------------------------------------------------------------ # async def _open_client( self, *, base_url: str, login: str, password: str, token: Optional[str] = None, ) -> ChattoClient: """Return a connected ``ChattoClient`` using token or login/password.""" if token: return ChattoClient(token=token, base_url=base_url) return await ChattoClient.login(login, password, base_url=base_url) async def connect(self, *, is_reconnect: bool = False) -> bool: """Connect to Chatto and start the realtime event stream. BasePlatformAdapter override """ logger.info("Chatto: connecting...") if not await self._ensure_token(): return False try: client = await self._require_client() except RuntimeError: self._set_fatal_error("connect_failed", "Chatto client not available", retryable=True) return False # Get our first own user info try: self.me = await client.me() except Exception as exc: logger.error("Chatto: failed to get user info: %s", exc) self._set_fatal_error( "chatto_auth_failed", f"Chatto auth failed: {exc}", retryable=False, ) try: assert(self._chatto_client) await self._chatto_client.close() finally: self._chatto_client = None return False # Announce online presence so the bot appears online in the member list. # The server treats this as a TTL, so _presence_refresh_loop below has to # keep re-announcing it — a single call here lapses back to offline. await self._announce_online() self._closing = False # Start background realtime WS event stream loop. self._ws_ready = asyncio.Event() self._ws_task = asyncio.create_task( self._chattolib_event_loop(), name="chatto-event-stream", ) self._presence_task = asyncio.create_task( self._presence_refresh_loop(), name="chatto-presence-refresh", ) self._mark_connected() logger.info( "Chatto: connected and authentiated to %s using login:'%s' (display_name:'%s' id: '%s')", self.chatto_config.base_url.value, self.me.login, self.me.display_name, self.me.id ) return True async def _announce_online(self) -> bool: """Tell the server we are online. Returns whether the call got through. Logged at warning level on failure: a silently dropped presence call is indistinguishable from a bot that is simply not running. """ try: client = await self._require_client() await client.update_presence(status=PresenceStatus.ONLINE) return True except Exception as exc: logger.warning("Chatto: presence refresh failed, bot may appear offline: %s", exc) return False async def _presence_refresh_loop(self) -> None: """Re-announce ONLINE until disconnect, since presence expires server-side. Failures are not fatal — the next tick tries again, so a blip in the presence endpoint costs at most one interval of visible offline time. """ while not self._closing: await self._sleep_interruptible(ChattoConstants.PRESENCE_REFRESH_INTERVAL) if self._closing: return await self._announce_online() async def disconnect(self) -> None: """Stop WebSocket, presence refresh, typing tasks, and clear state. BasePlatformAdapter override """ # No explicit offline broadcast: chattolib rejects OFFLINE outright # ("stop refreshing to go offline"), so cancelling the refresh loop # below is what actually takes the bot offline. self._ws_active = False self._closing = True # Cancel all typing tasks for chat_id in list(self._typing_tasks.keys()): await self.stop_typing(chat_id) if self._ws_task and not self._ws_task.done(): self._ws_task.cancel() try: await self._ws_task except (asyncio.CancelledError, Exception): pass self._ws_task = None if self._presence_task and not self._presence_task.done(): self._presence_task.cancel() try: await self._presence_task except (asyncio.CancelledError, Exception): pass self._presence_task = None if self._chatto_client: try: await self._chatto_client.close() except Exception: logger.exception("Chatto: error closing client") finally: self._chatto_client = None self._token = None logger.info("Chatto: disconnected") self._mark_disconnected() async def _seed_room(self, room_id: str) -> None: """Seed high-water mark from the newest events so a restart doesn't replay history.""" try: try: client = await self._require_client() timeline_page = await client.get_room_events(room_id) except RuntimeError: logger.debug("Chatto: _seed_room aborted - no client available for %s", room_id) return for ev in timeline_page.events: if ev.id: self._mark_seen(ev.id) logger.debug("Chatto: seeded room %s with %d events", room_id, len(timeline_page.events)) except Exception as e: logger.debug("Chatto: get room events failed for %s: %s", room_id, e) # ------------------------------------------------------------------ # # Realtime Event List # ------------------------------------------------------------------ # def _mark_seen(self, event_id: str) -> None: self._seen.append(event_id) while len(self._seen) > ChattoConstants.SEEN_CAP: self._seen.remove(self._seen[0]) # fastest removal of first item in a list. def _is_seen(self, event_id: str) -> bool: return event_id in self._seen # ------------------------------------------------------------------ # # WebSocket Realtime Transport # ------------------------------------------------------------------ # def _mentions_me(self, body: str) -> bool: """Whether the message @-mentions this bot, by login or display name.""" if not self.me: return False for handle in (self.me.login, self.me.display_name): if handle and f"@{handle}" in body: return True return False def _mentions_someone_else(self, body: str) -> bool: """Whether the message @-mentions a person who is not this bot. Broadcast handles are not a person — they address everyone present, the bot included, so they do not count as someone else. """ for handle in ChattoConstants.MENTION_RE.findall(body): if handle.lower() in ChattoConstants.BROADCAST_MENTIONS: continue if self.me and handle in (self.me.login, self.me.display_name): continue return True return False def _check_auth(self, user: User) -> bool: """We roll our own auth-systen on the against the chatto_config'ured allowed_users etc. because.. Hermes authz_mixin.py IS NOT SANE. """ if self.chatto_config.allow_all_users.value: return True if user.login in self.chatto_config.allowed_users.value: return True if user.id in self.chatto_config.allowed_users.value: return True logger.info("Chatto: rejecting message from unauthorized user '%s' (%s)", user.login, user.id) return False # ------------------------------------------------------------------ # # Inbound attachments # ------------------------------------------------------------------ # async def _download_attachment_bytes(self, url: str) -> bytes: """Download an attachment, refusing to buffer more than the gateway cap. The Content-Length header is checked first so an oversized asset is rejected before a single chunk is read; the running total is re-checked as chunks arrive, because a missing or lying header must not smuggle an unbounded body past the cap. """ import httpx max_bytes = get_inbound_media_max_bytes() chunks: List[bytes] = [] total = 0 async with httpx.AsyncClient( timeout=ChattoConstants.HTTP_TIMEOUT, follow_redirects=True, ) as http: async with http.stream("GET", url) as resp: resp.raise_for_status() declared = resp.headers.get("content-length") if declared: try: declared_size = int(declared) except ValueError: logger.debug("Chatto: ignoring invalid Content-Length %r", declared) else: validate_inbound_media_size( declared_size, media_type="attachment", max_bytes=max_bytes, ) async for chunk in resp.aiter_bytes(): total += len(chunk) validate_inbound_media_size( total, media_type="attachment", max_bytes=max_bytes, ) chunks.append(chunk) return b"".join(chunks) async def _cache_attachments( self, room_id: str, attachments: List[Any], ) -> Tuple[List[str], List[str], List[str]]: """Download message attachments into the gateway media cache. Returns ``(media_urls, media_types, media_kinds)`` — the paths are agent-visible cache paths, exactly what ``cache_media_bytes`` yields for every other platform. A failing attachment is logged and skipped: the message itself still reaches the agent. """ media_urls: List[str] = [] media_types: List[str] = [] media_kinds: List[str] = [] for att in attachments or []: asset_url = getattr(att, "asset_url", None) url = getattr(asset_url, "url", "") if asset_url else "" filename = getattr(att, "filename", "") or "" content_type = getattr(att, "content_type", "") or "" if not url: # Videos are announced before transcoding finishes, so the # signed URL can legitimately be missing on arrival. logger.info( "Chatto: attachment '%s' has no asset URL yet, skipping", filename, ) continue try: data = await self._download_attachment_bytes(url) cached = cache_media_bytes( data, filename=filename, mime_type=content_type, ) except Exception as e: logger.warning( "Chatto: failed to cache attachment '%s' (%s): %s", filename, content_type, e, ) continue if cached is None: logger.warning( "Chatto: attachment '%s' (%s) could not be cached, skipping", filename, content_type, ) continue media_urls.append(cached.path) media_types.append(cached.media_type) media_kinds.append(cached.kind) return media_urls, media_types, media_kinds @staticmethod def _message_type_for_media_kinds(media_kinds: List[str]) -> MessageType: """Pick the MessageType for a set of cached attachment kinds.""" if "document" in media_kinds: return MessageType.DOCUMENT if "image" in media_kinds: return MessageType.PHOTO if "video" in media_kinds: return MessageType.VIDEO if "audio" in media_kinds: return MessageType.AUDIO return MessageType.TEXT async def _dispatch_message_posted(self, payload: MessagePostedPayload) -> None: try: client = await self._require_client() except RuntimeError: logger.warning("Chatto: chattolib event loop aborted - no client available") return logger.info("Chatto WS: 'message_posted' event_payload:%s", payload) message = await payload.fetch_message(client=client) if message is None or message.deleted_at: return message_body = message.body or "" attachments = list(message.attachments or []) # A message carrying only an image/PDF has an empty body — dropping it # here is what made attachments sent to Hermes disappear silently. if not message_body and not attachments: return if message.actor_id in self._user_cache: # try the user cache. user = self._user_cache.get(message.actor_id) else: # get the user and update cache. directory_member = await client.get_user(user_id=message.actor_id) if directory_member is None: return user = directory_member.user if user is None: return self._user_cache[user.id] = user if user is None: return if not self._check_auth(user): return # Todo: use a function that either reads from cache or gets room kind again. if self._room_kinds.get(message.room_id) is None: room_viewer_state = await client.get_room(message.room_id) if room_viewer_state is None: return if room_viewer_state.room is None: return self._room_kinds[message.room_id] = room_viewer_state.room.kind or RoomKind.UNSPECIFIED room_kind = self._room_kinds.get(message.room_id) logger.info("message_body: %s room_kind: %s", message_body, room_kind) # require_mention deliberately gates channels only: in a channel the bot # is one of many listeners and must be addressed, whereas a DM is already # addressed at it — so DMs are always answered, mention or not. mentioned = False if (room_kind == RoomKind.CHANNEL and self.chatto_config.require_mention.value and self.me): if self.me.login and not mentioned: mentioned = bool(f"@{self.me.login}" in message_body) if self.me.display_name and not mentioned: mentioned = bool(f"@{self.me.display_name}" in message_body) if mentioned is False: logger.debug("Discarding message. Bot was not mentionend but require_mention is '%s'.", self.chatto_config.require_mention.value) return logger.info("mentioned: %s", mentioned) # With require_mention off we see every message in the channel, including # ones plainly aimed at a named colleague. Answering those would be # barging in, so acknowledge that we read it and stay quiet. Checked # after the bot-mention test above, so a message naming us *and* someone # else still counts as ours. if ( room_kind == RoomKind.CHANNEL and not self.chatto_config.require_mention.value and not self._mentions_me(message_body) and self._mentions_someone_else(message_body) ): logger.info("Chatto: message addresses someone else, acknowledging only") if self.chatto_config.reactions.value: await self.add_reaction(message.room_id, message.id, "🫥") return # Thread anchoring — if the incoming message is inside a thread, we # keep that thread by default; otherwise leave thread_id unset so # replies land at the root. thread_id = payload.thread_root_event_id or None if not thread_id and room_kind != RoomKind.DM and self.chatto_config.auto_thread.value: thread_id = message.id source = self.build_source( chat_id=payload.room_id, chat_name=self._room_names.get(message.room_id), chat_type=chat_type_for_room_kind(room_kind), user_id=message.actor_id, user_name=user.login, # use login, because display_name is changeable by anyone. thread_id=thread_id, message_id=payload.message_event_id, role_authorized=True, ) msg_type = MessageType.COMMAND if (message_body.lstrip().startswith("/")) else MessageType.TEXT my_body = message_body.lstrip() if msg_type == MessageType.COMMAND else message_body # Attachments — download and hand the local cache paths to the gateway, # which runs vision enrichment / document extraction off media_urls. media_urls, media_types, media_kinds = await self._cache_attachments( payload.room_id, attachments, ) if media_kinds and msg_type != MessageType.COMMAND: # Same precedence as the Teams/Signal adapters: document-context # injection gates strictly on DOCUMENT, image handling keys off the # per-path image/* MIME regardless of message_type. msg_type = self._message_type_for_media_kinds(media_kinds) message_event = MessageEvent( text=my_body, source=source, message_id=message.id, message_type=msg_type, media_urls=media_urls, media_types=media_types, timestamp=message.created_at or datetime.now(timezone.utc), raw_message=message, reply_to_message_id=message.id, reply_to_text=message_body, reply_to_author_id=user.id, reply_to_author_name=user.login, ) logger.info("Chatto: Dispatching MessageEvent to Hermes: %s", message_event) await self.handle_message(message_event) return async def _forward_reaction( self, event: RealtimeEvent, payload: ReactionPayload, *, removed: bool, ) -> None: """Forward a human reaction to the gateway's reaction hook surface. The handler is registered by the gateway via ``set_reaction_handler`` and fans out as ``reaction:added`` / ``reaction:removed`` through the HookRegistry. The dict shape mirrors the Slack adapter's — hook consumers are written against that contract, not against a per-platform one. Our own lifecycle reactions (👀/✅/❌) are dropped: forwarding them would feed the agent its own markers. """ actor_id = event.actor_id if actor_id and self.me and actor_id == self.me.id: return if not payload.room_id or not payload.message_event_id or not actor_id: return handler = getattr(self, "_reaction_handler", None) if handler is None: return action = "removed" if removed else "added" try: await handler( { "platform": ChattoConstants.PLATFORM_NAME, "event_name": f"reaction:{action}", "reaction": payload.emoji, "user_id": actor_id, "item_user_id": None, "item_type": "message", "channel_id": payload.room_id, "message_ts": payload.message_event_id, "event_ts": event.id, "raw_event": event, } ) except Exception: # pragma: no cover - the hook contract is non-blocking logger.debug("Chatto: reaction hook forwarding failed", exc_info=True) async def _handle_realtime_event(self, event: RealtimeEvent) -> None: if self._is_seen(event.id): return if event.actor_id is None: return logger.info("EVENT happened: '%s' from %s", event.kind, event.actor_id) if (event_payload := event.get("message_posted")) is not None: # Self-event filter — the actor_id on the envelope is authoritative # (chattolib does NOT filter this itself; see chatto-bridge notes). actor_id = event.actor_id if actor_id and self.me and actor_id == self.me.id: return await self._dispatch_message_posted(event_payload) elif event.kind in ("reaction_added", "reaction_removed"): reaction_payload = event.get(event.kind) if reaction_payload is not None: await self._forward_reaction( event, cast(ReactionPayload, reaction_payload), removed=event.kind == "reaction_removed", ) # confirmed: elif event.kind in ("mention_notification", "projection_event", "presence_changed", "notification_dismissed", "room_marked_as_read", "user_typing", "notification_created", "new_direct_message_notification", "message_edited", "message_retracted"): logger.debug("Chatto: '%s' event received. Not yet implemented or not needed.", event.kind) else: logger.error("Chatto: unknown event kind: '%s'", event.kind) async def _chattolib_event_loop(self) -> None: """Event loop using chattolib's stream_events. This replaces the manual WebSocket loop with chattolib's high-level stream_events() which provides pre-decoded RealtimeEvent objects. """ delay = ChattoConstants.WS_RECONNECT_INITIAL_BACKOFF while not self._closing: try: client = await self._require_client() await self._refresh_rooms() logger.info("Chatto: starting chattolib event stream with %d rooms", len(self._watch_room_ids)) async for event in stream_events(client): if self._closing: return await self._handle_realtime_event(event) # Iterator exited cleanly — treat as a normal close and reconnect # with the local backoff (no server hint available). logger.info("Chatto: realtime stream ended, reconnecting") except asyncio.CancelledError: return except ChattoRealtimeCloseError as exc: if not exc.reconnect: logger.error( "Chatto: realtime closed by server (%s: %s), not reconnecting", exc.code, exc.message, ) return wait = max(exc.retry_after_ms / 1000.0, ChattoConstants.WS_RECONNECT_INITIAL_BACKOFF) logger.warning( "Chatto: realtime closed by server (%s), reconnecting in %.1fs", exc.code, wait, ) delay = ChattoConstants.WS_RECONNECT_INITIAL_BACKOFF # server hint supersedes local backoff await self._sleep_interruptible(wait) continue except ChattoRealtimeError as exc: if getattr(exc, "fatal", False): logger.error("Chatto: fatal realtime error (%s): %s", exc.code, exc.message) return logger.warning( "Chatto: realtime error (%s: %s), reconnecting in %.1fs", exc.code, exc.message, delay, ) except Exception as exc: logger.warning( "Chatto: unexpected realtime error: %s, reconnecting in %.1fs", exc, delay, ) if self._closing: return jitter = delay * 0.2 * random.random() await self._sleep_interruptible(delay + jitter) delay = min(delay * 2, ChattoConstants.WS_RECONNECT_MAX_BACKOFF) async def _sleep_interruptible(self, seconds: float) -> None: """Sleep in short slices so disconnect() cancels promptly.""" end = asyncio.get_running_loop().time() + seconds while not self._closing: remaining = end - asyncio.get_running_loop().time() if remaining <= 0: return await asyncio.sleep(min(remaining, 0.5)) async def _refresh_rooms(self) -> None: """Refresh room list via ConnectRPC, join and seed any newly discovered rooms.""" try: client = await self._require_client() except RuntimeError: logger.warning("Chatto WS: _refresh_rooms aborted - no client available") return try: rooms_list = await client.list_rooms() new_room_ids: List[str] = [] for room_with_state in rooms_list: if not room_with_state: continue room_obj = room_with_state.room or None if not room_obj: continue self._room_names[room_obj.id] = room_obj.name self._room_kinds[room_obj.id] = room_obj.kind if room_with_state.viewer_state.is_member and room_obj.id not in self._watch_room_ids: new_room_ids.append(room_obj.id) if not new_room_ids: return logger.info("Chatto WS: discovered %d new room(s): %s", len(new_room_ids), new_room_ids) for rid in new_room_ids: if self._room_kinds.get(rid) != RoomKind.DM: await client.join_room(rid) # but list_rooms() would not return any if we were not already joined?! await self._seed_room(rid) self._watch_room_ids.append(rid) watch_room_names: list[str] = [] for rid in self._watch_room_ids: watch_room_names.append(self._room_names[rid] + " (" + rid + ")") logger.info("Chatto WS: Watching %d room(s): %s", len(self._watch_room_ids), ", ".join(watch_room_names)) except Exception: logger.warning("Chatto WS: _refresh_rooms failed", exc_info=True) # ------------------------------------------------------------------ # # Read state & notification dismissal (best-effort, Chatto-unique) # ------------------------------------------------------------------ # # Best-effort: mark all watched rooms as read (room_id may be undefined here) for _rid in list(self._watch_room_ids): try: await client.mark_room_as_read(room_id=_rid) await client.dismiss_all_notifications() except Exception: logger.debug("Chatto: mark_room_as_read failed for %s", _rid, exc_info=True) # ------------------------------------------------------------------ # # Sending (ConnectRPC — unchanged from polling version) # ------------------------------------------------------------------ # async def send( self, chat_id: str, content: str, reply_to: Optional[str] = None, metadata: Optional[Dict[str, Any]] = None, ) -> SendResult: """Send a message to a Chatto room. Long messages are split into chunks via ``truncate_message`` and each chunk is sent as a separate CreateMessage call. The first chunk's message ID is returned as ``message_id``. When ``auto_thread`` is enabled and the incoming message was a regular room message (not already in a thread), the first chunk is sent as a room message and its ID becomes the thread root. Subsequent chunks are sent in that thread. This mirrors Discord's auto_thread behavior. BasePlatformAdapter override """ if not content: return SendResult(success=False, error="Empty message") formatted = self.format_message(content) chunks = self.truncate_message(formatted, ChattoConstants.MAX_MESSAGE_LENGTH) # Thread support — resolve thread_id once # DM rooms don't support threads, so skip threading for DMs thread_id = (metadata or {}).get("thread_id") # Only use reply_to as thread_id if auto_thread is enabled. # When auto_thread=false, responses go directly in the room # without threading under the incoming message. if reply_to and self.chatto_config.auto_thread.value: # reply_to might be the incoming message ID. If we already have # thread_id from metadata, keep it (it's the thread root). # Only use reply_to as thread_id if we don't already have one. if not thread_id: thread_id = reply_to # Check if this is a DM room — DMs don't support threads room_kind = self._room_kinds.get(chat_id) is_dm = room_kind == RoomKind.DM if is_dm: thread_id = None # Auto-thread: by default, Chatto creates a thread for replies to room # messages (not DMs, not already in a thread). This keeps conversations # organized in the room. Can be disabled via extra.auto_thread=false. use_auto_thread = self.chatto_config.auto_thread.value and not thread_id and not is_dm message_ids: List[str] = [] last_resp: Optional[Any] = None last_error: Optional[str] = None retryable = False try: client = await self._require_client() except RuntimeError: return SendResult(success=False, error="Chatto client not available", retryable=True) for i, chunk in enumerate(chunks): try: msg_obj = await client.post_message( room_id=chat_id, body=chunk, thread_root_event_id=str(thread_id) if thread_id else "", ) except ChattoError as e: last_error = str(e) retryable = True break except Exception as e: last_error = str(e) retryable = True break last_resp = msg_obj self._mark_seen(msg_obj.id) message_ids.append(msg_obj.id) self._our_message_ids.add(msg_obj.id) # If we sent a message WITHOUT a thread_id, this message could # become a thread root if someone replies to it if not thread_id: self._our_thread_roots.add(msg_obj.id) # Auto-thread: first chunk becomes the thread root, # subsequent chunks go in the thread if use_auto_thread and i == 0 and not thread_id: thread_id = msg_obj.id # Nothing got through at all — report the failure instead of a phantom success. if not message_ids: return SendResult( success=False, error=last_error or "Chatto: message could not be sent", retryable=retryable, ) first_id = message_ids[0] # ------------------------------------------------------------------ # # Thread following (best-effort, Chatto-unique) # ------------------------------------------------------------------ # if thread_id: try: await client.follow_thread(chat_id, thread_id) except Exception: logger.debug("Chatto: follow_thread failed for %s/%s", chat_id, thread_id, exc_info=True) # A later chunk failed after earlier ones went out: partial delivery. if last_error: logger.warning( "Chatto: sent %d/%d chunk(s) to %s before failing: %s", len(message_ids), len(chunks), chat_id, last_error, ) return SendResult(success=True, message_id=first_id, raw_response=last_resp) def format_message(self, content: str) -> str: """Normalise outgoing text for Chatto. Chatto renders Markdown natively, so there is nothing to escape or translate — the only transformations here are the ones that measurably render wrong: CRLF line endings (which show up as stray blank lines) and runs of more than two blank lines. BasePlatformAdapter override """ if not content: return content normalised = content.replace("\r\n", "\n").replace("\r", "\n") while "\n\n\n\n" in normalised: normalised = normalised.replace("\n\n\n\n", "\n\n\n") return normalised async def edit_message( self, chat_id: str, message_id: str, content: str, *, finalize: bool = False, ) -> SendResult: """Edit a message we previously sent, via MessageService/UpdateMessage. The stream consumer drives streaming replies through this: without the override the base class reports "Not supported" and every incremental update arrives as a *new* message. ``finalize`` is a no-op for Chatto — an edit is an edit here, there is no in-progress card state to close out (hence no ``REQUIRES_EDIT_FINALIZE``). Content that exceeds the per-message limit is refused rather than silently truncated, so the caller falls back to ``send()``, which splits across messages. BasePlatformAdapter override """ if not message_id: return SendResult(success=False, error="Chatto: no message id to edit") if not content: return SendResult(success=False, error="Empty message") formatted = self.format_message(content) if len(formatted) > ChattoConstants.MAX_MESSAGE_LENGTH: # Refuse instead of truncating: the caller's fallback path splits. return SendResult( success=False, error=( f"Chatto: edit exceeds {ChattoConstants.MAX_MESSAGE_LENGTH} " f"chars ({len(formatted)})" ), ) try: client = await self._require_client() except RuntimeError: return SendResult(success=False, error="Chatto client not available", retryable=True) try: msg = await client.update_message( room_id=str(chat_id), event_id=str(message_id), body=formatted, ) except ChattoError as e: logger.warning("Chatto: UpdateMessage failed for %s: %s", message_id, e) return SendResult(success=False, error=str(e), retryable=True) except Exception as e: logger.warning("Chatto: UpdateMessage error for %s: %s", message_id, e) return SendResult(success=False, error=str(e), retryable=False) # Our own edit comes back as a message_edited event; mark it seen so it # is never mistaken for inbound traffic. edited_id = getattr(msg, "id", "") or str(message_id) self._mark_seen(edited_id) return SendResult(success=True, message_id=edited_id, raw_response=msg) async def delete_message(self, chat_id: str, message_id: str) -> bool: """Delete a message via MessageService/DeleteMessage. Used by the stream consumer's fresh-final cleanup (removing a preview message once the completed reply has been sent) and by the ephemeral reply TTL. BasePlatformAdapter override """ if not chat_id or not message_id: return False try: client = await self._require_client() except RuntimeError: logger.warning("Chatto: DeleteMessage — client unavailable") return False try: return bool(await client.delete_message( room_id=str(chat_id), event_id=str(message_id), )) except ChattoError as e: logger.warning("Chatto: DeleteMessage failed for %s: %s", message_id, e) return False except Exception as e: logger.warning("Chatto: DeleteMessage error for %s: %s", message_id, e) return False async def create_handoff_thread( self, parent_chat_id: str, name: str, ) -> Optional[str]: """Anchor a session handoff in a fresh thread under *parent_chat_id*. Chatto threads hang off a message, not off the room, so we post a seed message and hand its ID back as the thread root — the same shape the Slack adapter uses. DMs don't support threads, so they get ``None`` and the watcher keeps delivering into the DM itself. BasePlatformAdapter override """ if not parent_chat_id: return None if self._room_kinds.get(parent_chat_id) == RoomKind.DM: logger.debug("Chatto: handoff thread skipped — %s is a DM", parent_chat_id) return None try: client = await self._require_client() except RuntimeError: logger.warning("Chatto: handoff thread — client unavailable") return None seed_text = f"🧵 Hermes handoff — **{(name or 'session').strip()[:80]}**" try: msg = await client.post_message(room_id=str(parent_chat_id), body=seed_text) except Exception as e: logger.warning( "Chatto: handoff thread seed-post failed for room %s: %s", parent_chat_id, e, ) return None seed_id = getattr(msg, "id", "") or "" if not seed_id: logger.warning("Chatto: handoff thread seed-post returned no message id") return None self._mark_seen(seed_id) self._our_message_ids.add(seed_id) self._our_thread_roots.add(seed_id) try: await client.follow_thread(str(parent_chat_id), seed_id) except Exception: logger.debug( "Chatto: follow_thread failed for handoff %s/%s", parent_chat_id, seed_id, exc_info=True, ) return seed_id # Overridden from BaseAdapter: async def send_typing(self, chat_id: str, metadata=None) -> None: """Start a persistent typing indicator for a room. Sends a typing ping every 10 seconds (Chatto's indicator likely lasts ~8-10s). The background loop runs until ``stop_typing()`` is called or the task is cancelled. BasePlatformAdapter override """ if chat_id in self._typing_tasks: return # already running async def _typing_loop() -> None: try: while True: try: try: client = await self._require_client() except RuntimeError: return await client.update_typing_indicator(room_id=str(chat_id)) except asyncio.CancelledError: return except Exception: pass await asyncio.sleep(10) except asyncio.CancelledError: pass finally: self._typing_tasks.pop(chat_id, None) self._typing_tasks[chat_id] = asyncio.create_task(_typing_loop()) async def stop_typing(self, chat_id: str) -> None: """Stop the persistent typing indicator for a room. BasePlatformAdapter override """ task = self._typing_tasks.pop(chat_id, None) if task: task.cancel() try: await task except (asyncio.CancelledError, Exception): pass async def get_chat_info(self, chat_id: str) -> Dict[str, Any]: """Get information about a chat/room. BasePlatformAdapter override """ name = self._room_names.get(chat_id, chat_id) kind = self._room_kinds.get(chat_id) return { "name": name, "type": chat_type_for_room_kind(kind).value, } # ------------------------------------------------------------------ # # Reactions # ------------------------------------------------------------------ # @staticmethod def _emoji_to_shortcode(emoji: str) -> str: """Convert a unicode emoji to a Chatto shortcode name. If the emoji is already a shortcode (no unicode mapping found), return it as-is. """ shortcode = ChattoConstants.EMOJI_TO_SHORTCODE.get(emoji) if shortcode: return shortcode # Already a shortcode like "thumbsup" — return as-is return emoji async def add_reaction(self, room_id: str, message_id: str, emoji: str) -> bool: """Add a reaction to a message via MessageService/AddReaction.""" shortcode = self._emoji_to_shortcode(emoji) try: try: client = await self._require_client() except RuntimeError: logger.warning("Chatto: AddReaction — client unavailable") return False result = await client.add_reaction( room_id=room_id, message_event_id=message_id, emoji=shortcode, ) return result except ChattoError as e: logger.warning("Chatto: AddReaction failed: %s", e) return False except Exception as e: logger.warning("Chatto: AddReaction error: %s", e) return False async def remove_reaction(self, chat_id: str, message_id: str, emoji: str) -> bool: """Remove a reaction from a message via MessageService/RemoveReaction.""" shortcode = self._emoji_to_shortcode(emoji) try: try: client = await self._require_client() except RuntimeError: logger.warning("Chatto: RemoveReaction — client unavailable") return False result = await client.remove_reaction( room_id=str(chat_id), message_event_id=str(message_id), emoji=shortcode, ) return result except ChattoError as e: logger.warning("Chatto: RemoveReaction failed: %s", e) return False except Exception as e: logger.warning("Chatto: RemoveReaction error: %s", e) return False # ------------------------------------------------------------------ # # DM initiation (Chatto-unique) # ------------------------------------------------------------------ # async def start_dm(self, user_id: str) -> Optional[str]: """Start a direct message with a user via RoomService/StartDM. Returns the room ID on success, or None on failure. """ if not user_id: return None try: try: client = await self._require_client() except RuntimeError: return None room = await client.start_dm(participant_ids=[str(user_id)]) self._room_names[room.id] = room.name self._room_kinds[room.id] = room.kind return room.id except ChattoError as e: logger.debug("Chatto: StartDM failed: %s", e) return None except Exception as e: logger.debug("Chatto: StartDM error: %s", e) return None # ------------------------------------------------------------------ # # Room creation (Chatto-unique) # ------------------------------------------------------------------ # async def create_room( self, name: str, description: str = "", group_id: str = "", universal: bool = True, ) -> Optional[str]: """Create an ad-hoc room via RoomService/CreateRoom. Returns the room ID on success, or None on failure. """ try: try: client = await self._require_client() except RuntimeError: return None room = await client.create_room( name=name, group_id=group_id or "", description=description, universal=universal, ) rid = str(room.id) if room else "" if rid: self._room_names[rid] = room.name self._room_kinds[rid] = room.kind return rid logger.debug("Chatto: CreateRoom returned no room id") return None except ChattoError as e: logger.debug("Chatto: CreateRoom failed: %s", e) return None except Exception as e: logger.debug("Chatto: CreateRoom error: %s", e) return None # ------------------------------------------------------------------ # # Processing lifecycle hooks (reactions-based, like Discord) # ------------------------------------------------------------------ # def _event_room_and_message_id(self, event: MessageEvent) -> Tuple[str, str]: """Extract room_id and message_id from a MessageEvent.""" message_id = event.message_id or "" chat_id = event.source.chat_id return chat_id, message_id async def on_processing_start(self, event: MessageEvent) -> None: """Add an 👀 (eyes) reaction to the incoming message. BasePlatformAdapter override """ if not self.chatto_config.reactions.value: return chat_id, message_id = self._event_room_and_message_id(event) if not chat_id or not message_id: # Routine, not a fault: the gateway runs agent-initiated turns # (heartbeat polls, goal continuations) through the same pipeline # with message_id=None, and there is no inbound message to mark. logger.debug( "Chatto: nothing to react to (chat_id=%r, message_id=%r)", chat_id, message_id, ) return await self.add_reaction(chat_id, message_id, "👀") async def on_processing_complete( self, event: MessageEvent, outcome: ProcessingOutcome ) -> None: """Swap the 👀 reaction for ✅ (success) or ❌ (failure). BasePlatformAdapter override """ if not self.chatto_config.reactions.value: return chat_id, message_id = self._event_room_and_message_id(event) if not chat_id or not message_id: return # Remove the processing eyes reaction await self.remove_reaction(chat_id, message_id, "👀") # Add the outcome reaction if outcome == ProcessingOutcome.SUCCESS: await self.add_reaction(chat_id, message_id, "✅") elif outcome == ProcessingOutcome.FAILURE: await self.add_reaction(chat_id, message_id, "❌") # ------------------------------------------------------------------ # # Asset upload (chunked) # ------------------------------------------------------------------ # async def _upload_asset(self, room_id: str, file_path: str) -> Optional[str]: """Upload a file via the chunked AssetUploadService. Returns the asset ID on success, or None on failure. """ try: with open(file_path, "rb") as f: file_data = f.read() except Exception as e: logger.error("Chatto: failed to read file %s — %s", file_path, e) return None if not file_data: logger.error("Chatto: file %s is empty", file_path) return None file_size = len(file_data) file_name = os.path.basename(file_path) mime_type = mimetypes.guess_type(file_path)[0] or "application/octet-stream" sha256_hash = hashlib.sha256(file_data).hexdigest() try: try: client = await self._require_client() except RuntimeError: logger.error("Chatto: upload aborted - no client available") return None # Step 1: Create upload session upload = await client.create_upload( room_id=room_id, filename=file_name, size=file_size, sha256=sha256_hash, content_type=mime_type, ) # AssetUpload names this upload_id, not id — reading it through an # untyped getattr default is what let the mismatch reach production. upload_id = upload.upload_id if not upload_id: logger.error("Chatto: CreateUpload returned no upload ID") return None # Step 2: Upload chunks offset = 0 while offset < file_size: chunk = file_data[offset:offset + ChattoConstants.UPLOAD_CHUNK_SIZE] chunk_sha256 = hashlib.sha256(chunk).hexdigest() await client.upload_chunk( upload_id=upload_id, offset=offset, content=chunk, chunk_sha256=chunk_sha256, ) offset += len(chunk) # Step 3: Complete upload upload, asset = await client.complete_upload(upload_id=upload_id) if not asset: logger.error("Chatto: CompleteUpload returned no asset") return None asset_id = str(getattr(cast(Any, asset), "id", "")) logger.info("Chatto: uploaded %s as asset %s (%d bytes)", file_name, asset_id, file_size) return asset_id except ChattoError as e: logger.error("Chatto: upload failed: %s", e) return None except Exception as e: logger.error("Chatto: upload error: %s", e) return None async def _post_attachment_message( self, chat_id: str, asset_ids: List[str], caption: Optional[str], reply_to: Optional[str], metadata: Optional[Dict[str, Any]], ) -> SendResult: """Post one message carrying already-uploaded assets.""" thread_id = (metadata or {}).get("thread_id") if reply_to: thread_id = reply_to try: try: client = await self._require_client() except RuntimeError: return SendResult(success=False, error="Chatto client not available", retryable=True) msg = await client.post_message( room_id=str(chat_id), body=self.format_message(caption) if caption else "", attachment_asset_ids=asset_ids, thread_root_event_id=str(thread_id) if thread_id else "", ) self._mark_seen(msg.id) self._our_message_ids.add(msg.id) return SendResult(success=True, message_id=msg.id, raw_response=msg) except ChattoError as e: return SendResult(success=False, error=str(e), retryable=True) except Exception as e: return SendResult(success=False, error=str(e), retryable=False) async def _send_local_attachment( self, chat_id: str, file_path: str, caption: Optional[str], reply_to: Optional[str], metadata: Optional[Dict[str, Any]], *, kind: str, ) -> SendResult: """Upload a local file and post it as a native Chatto attachment. Shared by ``send_image_file``/``send_document``/``send_video``/ ``send_voice`` — the upload mechanics are identical, only the wording of the failure notice differs. On failure we send that notice as text and never the host path (it leaks the Hermes home layout). """ notice = f"⚠️ Couldn't deliver the {kind} attachment." safe_path = self.validate_media_delivery_path(file_path) if not safe_path: logger.warning( "[%s] send %s: unsafe path %s", self.name, kind, file_path, ) text = f"{caption}\n{notice}" if caption else notice return await self.send(chat_id, text, reply_to=reply_to, metadata=metadata) asset_id = await self._upload_asset(str(chat_id), safe_path) if not asset_id: logger.warning( "[%s] send %s: upload failed for %s", self.name, kind, safe_path, ) text = f"{caption}\n{notice}" if caption else notice return await self.send(chat_id, text, reply_to=reply_to, metadata=metadata) return await self._post_attachment_message( chat_id, [asset_id], caption, reply_to, metadata, ) async def send_image_file( self, chat_id: str, image_path: str, caption: Optional[str] = None, reply_to: Optional[str] = None, metadata: Optional[Dict[str, Any]] = None, **kwargs, ) -> SendResult: """Send a local image file via the chunked upload API. The parameter is ``image_path``, not ``file_path``: every caller passes it by keyword (``gateway/run.py:22354``, ``:22470``, and the base class's own ``send_multiple_images`` file:// branch), so a renamed parameter makes each of those raise TypeError and silently degrade to a text notice. BasePlatformAdapter override """ return await self._send_local_attachment( chat_id, image_path, caption, reply_to, metadata, kind="image", ) async def send_document( self, chat_id: str, file_path: str, caption: Optional[str] = None, file_name: Optional[str] = None, reply_to: Optional[str] = None, metadata: Optional[Dict[str, Any]] = None, **kwargs, ) -> SendResult: """Send a local file as a native Chatto attachment. ``file_name`` is the user-facing name the agent chose; Chatto takes the filename from the upload session, so it only matters for the failure notice. BasePlatformAdapter override """ result = await self._send_local_attachment( chat_id, file_path, caption, reply_to, metadata, kind="file", ) if not result.success and file_name: logger.debug("Chatto: document delivery failed for %s", file_name) return result async def send_video( self, chat_id: str, video_path: str, caption: Optional[str] = None, reply_to: Optional[str] = None, metadata: Optional[Dict[str, Any]] = None, **kwargs, ) -> SendResult: """Send a local video as a native Chatto attachment (Chatto transcodes and plays it inline). BasePlatformAdapter override """ return await self._send_local_attachment( chat_id, video_path, caption, reply_to, metadata, kind="video", ) async def send_voice( self, chat_id: str, audio_path: str, caption: Optional[str] = None, reply_to: Optional[str] = None, metadata: Optional[Dict[str, Any]] = None, **kwargs, ) -> SendResult: """Send a local audio file as a native Chatto attachment. Chatto has no dedicated voice-bubble type, so this is an ordinary audio attachment — still far better than the base class's text notice. BasePlatformAdapter override """ return await self._send_local_attachment( chat_id, audio_path, caption, reply_to, metadata, kind="audio", ) async def send_image( self, chat_id: str, image_url: str, caption: Optional[str] = None, reply_to: Optional[str] = None, metadata: Optional[Dict[str, Any]] = None, ) -> SendResult: """Send an image to a Chatto room. Tries to download the image from the URL and upload it as a native attachment. Falls back to sending the URL as a link (Chatto renders link previews) if the download fails. BasePlatformAdapter override """ # Try downloading and uploading as attachment try: import tempfile import urllib.request as _urllib_request # Download to a temp file parsed = urlsplit(image_url) url_path = parsed.path ext = os.path.splitext(url_path)[1] or ".png" tmp_fd, tmp_path = tempfile.mkstemp(suffix=ext, prefix="chatto_img_") try: os.close(tmp_fd) req = _urllib_request.Request(image_url, headers={"User-Agent": "Hermes/1.0"}) try: import ssl ctx = ssl.create_default_context() except Exception: ctx = None with _urllib_request.urlopen(req, timeout=ChattoConstants.HTTP_TIMEOUT, context=ctx) as resp: with open(tmp_path, "wb") as f: f.write(resp.read()) # Upload as attachment result = await self.send_image_file( chat_id, tmp_path, caption=caption, reply_to=reply_to, metadata=metadata, ) if result.success: return result finally: try: os.unlink(tmp_path) except OSError: pass except Exception as e: logger.debug("Chatto: send_image download/upload failed, falling back to link: %s", e) # Fallback: send as link (Chatto renders link previews) text = image_url if caption: text = f"{caption}\n{image_url}" return await self.send(chat_id, text, reply_to=reply_to, metadata=metadata) async def _materialise_image(self, image_url: str) -> Tuple[Optional[str], bool]: """Resolve one ``send_multiple_images`` entry to a local file path. Accepts ``http(s)://`` URLs (downloaded to a temp file), ``file://`` URIs and bare paths. Returns ``(path, is_temp)`` — the caller unlinks when ``is_temp``. ``(None, False)`` means the entry is unusable. """ import tempfile from urllib.parse import unquote as _unquote if image_url.startswith(("http://", "https://")): parsed = urlsplit(image_url) ext = os.path.splitext(parsed.path)[1] or ".png" tmp_fd, tmp_path = tempfile.mkstemp(suffix=ext, prefix="chatto_img_") os.close(tmp_fd) try: data = await self._download_attachment_bytes(image_url) with open(tmp_path, "wb") as f: f.write(data) except Exception as e: logger.warning("Chatto: image download failed for %s: %s", image_url, e) try: os.unlink(tmp_path) except OSError: pass return None, False return tmp_path, True local = image_url if local.startswith("file://"): local = _unquote(urlsplit(local).path) return self.validate_media_delivery_path(local), False async def send_multiple_images( self, chat_id: str, images: List[Tuple[str, str]], metadata: Optional[Dict[str, Any]] = None, human_delay: float = 0.0, ) -> None: """Send a batch of images as ONE message with several attachments. The base implementation posts each image separately; a Chatto message carries a list of attachment assets, so a batch belongs in a single message (and a single notification). ``human_delay`` is ignored deliberately — there is only one outbound call to pace. Entries that can't be fetched are dropped with a warning; if nothing survives, we fall back to the base class so the user still gets the links. BasePlatformAdapter override """ if len(images or []) < 2: await super().send_multiple_images( chat_id, images, metadata=metadata, human_delay=human_delay, ) return asset_ids: List[str] = [] captions: List[str] = [] for image_url, alt_text in images: path, is_temp = await self._materialise_image(image_url) if not path: logger.warning("Chatto: skipping unusable image %s", image_url) continue try: asset_id = await self._upload_asset(str(chat_id), path) finally: if is_temp: try: os.unlink(path) except OSError: pass if not asset_id: logger.warning("Chatto: upload failed for image %s", image_url) continue asset_ids.append(asset_id) if alt_text: captions.append(alt_text) if not asset_ids: logger.warning( "Chatto: no image survived upload, falling back to per-image delivery", ) await super().send_multiple_images( chat_id, images, metadata=metadata, human_delay=human_delay, ) return if len(asset_ids) < len(images): logger.warning( "Chatto: sending %d of %d images — the rest could not be uploaded", len(asset_ids), len(images), ) await self._post_attachment_message( chat_id, asset_ids, "\n".join(captions) or None, None, metadata, ) # --------------------------------------------------------------------------- # Cron / out-of-process delivery # --------------------------------------------------------------------------- async def hermes_standalone_sender_fn( pconfig, chat_id, message, *, thread_id=None, media_files=None, force_document=False, ) -> SendResult: """Deliver a message to Chatto without a running gateway adapter. Do not modify signature. Used by cron / scheduled routines that run out-of-process. Creates a short-lived chattolib client, posts, and closes. """ chatto_config: ChattoConfiguration = ChattoConfiguration(pconfig=pconfig) # Create a temporary client for standalone sending — we need a base URL plus # either a token or a full login/password pair. has_credentials = bool( chatto_config.token.value or (chatto_config.login.value and chatto_config.password.value) ) if not chatto_config.base_url.value or not has_credentials: return SendResult(success=False, error="Chatto: base URL or credentials missing") client: ChattoClient try: if chatto_config.token.value: client = ChattoClient(base_url=chatto_config.base_url.value, token=chatto_config.token.value) else: client = await ChattoClient.login( base_url=chatto_config.base_url.value, login=chatto_config.login.value, password=chatto_config.password.value, ) except Exception as exc: return SendResult(success=False, error=f"Chatto login failed: {exc}") try: kwargs: Dict[str, Any] = {} if chatto_config.auto_thread.value and thread_id: kwargs["in_reply_to"] = thread_id if media_files and media_files.get("attachment_asset_ids"): kwargs["attachment_asset_ids"] = list(media_files["attachment_asset_ids"]) try: posted = await client.post_message(chat_id, message, **kwargs) except Exception as exc: return SendResult(success=False, error=str(exc)) return SendResult(success=True, message_id=getattr(posted, "id", "") or None) finally: try: await client.close() except Exception as exc: logger.error( "Chatto standalone: error closing short-lived client (perhaps already closed): %s", exc, ) def hermes_validate_config(config: PlatformConfig) -> bool: """" Function name should be the same as register argument name with "hermes_" prefix, so we know that it is needed for plugin register(). Do not change signature. - config Check whether Chatto Plugin is configured. Compare to hermes_is_connected().""" chatto_config = ChattoConfiguration(pconfig=config) if len(chatto_config.allowed_users.value) > 0 and chatto_config.allow_all_users.value: logger.info("Chatto: Conflicting configuration. Either use 'allowed_users' or 'allow_all_users' but not both.") return False if chatto_config.base_url.value: if (chatto_config.token.value is not None) or (chatto_config.login.value and chatto_config.password.value): return True else: logger.error("Chatto: Minimally, either token or login/password must be set.") else: logger.error("Chatto: base_url must be set.") return False def hermes_check_fn() -> bool: """Check if Chatto is configured and dependencies are available. Add real logic?! Or just .. there are no dependencies.. always return true. Really. Docs suck.""" try: import chattolib.client # noqa: F401 — vendored dependency probe return True except ImportError: return False # --------------------------------------------------------------------------- # is_connected probe # --------------------------------------------------------------------------- def hermes_is_connected(config: PlatformConfig) -> bool: """Check whether Chatto Plugin is connected. But where to? To the Hermes Agent? To Chatto Server? The Hermes Agent plugin docs suck and it seems there are many functions to do the same.""" return bool(hermes_validate_config(config) and config.enabled) def hermes_setup_fn() -> None: """Interactive setup wizard for Chatto. Is called by and only works in Hermes CLI context. Function name should be the same as register argument name with "hermes_" prefix, so we know that it is needed for plugin register(). """ from hermes_cli.setup import ( prompt, prompt_yes_no, save_env_value, get_env_value, print_header, print_info, print_warning, print_success, ) url = prompt( "Chatto server URL (e.g. https://chat.example.com) or leave blank for default ChattoHQ on chat.chatto.run:") if url: save_env_value(ChattoConfiguration.base_url.env_name, url) login = prompt("Chatto login (username):") if login: save_env_value(ChattoConfiguration.login.env_name, login) password = prompt("Chatto password:", password=True) if password: save_env_value(ChattoConfiguration.password.env_name, password) channels = prompt("Room IDs to watch (comma-separated, or empty for all):") if channels: save_env_value(ChattoConfiguration.channels_list.env_name, channels) home = prompt("Home room ID for notifications (or empty):") if home: save_env_value(ChattoConfiguration.home_channel.env_name, home) allow_all = prompt_yes_no("Allow all users to talk? (true/false):") if allow_all: save_env_value(ChattoConfiguration.allow_all_users.env_name, str(allow_all)) print_success("\n✓ Chatto configured. Restart the gateway to activate.") def hermes_env_enablement_fn() -> Optional[dict]: """Seed PlatformConfig.extra from env vars. Returns a dict compatible with the PlatformConfig merge hook (or None when no env-provided values are present). Called by the platform registry during load_gateway_config(). Return None when the platform isn't minimally configured — the caller then skips auto-enabling. Return a dict to seed extras. The special 'home_channel' key is extracted and becomes a proper HomeChannel dataclass on the PlatformConfig; every other key is merged into PlatformConfig.extra. Function name should be the same as register argument name with "hermes_" prefix, so we know that it is needed for plugin register(). """ # Seed keys must be the config.yaml "extra" keys (config_key), NOT the env # var names — ChattoConfiguration reads extra[config_key]. seed: Dict[str, Any] = { ChattoConfiguration.base_url.config_key: ( os.getenv(ChattoConfiguration.base_url.env_name) or ChattoClient.DEFAULT_BASE_URL ).strip(), } for field in ChattoConfiguration.fields(): if field.field_name == ChattoConfiguration.base_url.field_name: continue env_value = os.getenv(field.env_name) if env_value: seed[field.config_key] = env_value.strip() logger.debug("seed: %s", {k: v for k, v in seed.items() if k not in ("token", "password")}) return seed # --------------------------------------------------------------------------- # Plugin registration entry point # --------------------------------------------------------------------------- # What each capability is called in the startup banner, keyed by the method # that implements it. Derived from real overrides rather than hard-coded, so # dropping a method drops its claim from the log instead of leaving a lie. _CAPABILITY_LABELS = { "send": "text", "send_image_file": "images", "send_multiple_images": "image batches (bundled into one message)", "send_video": "video", "send_voice": "voice messages", "send_document": "documents", "add_reaction": "reactions", "edit_message": "message editing", "delete_message": "message deletion", "send_typing": "typing indicators", "create_handoff_thread": "threads", "start_dm": "direct messages", "create_room": "room creation", } def _capabilities() -> List[str]: """Name the things this adapter genuinely implements itself. A capability counts only when ChattoAdapter overrides the base method — inheriting BasePlatformAdapter's fallback means the feature is not natively supported, and announcing it would mislead. """ found = [ label for name, label in _CAPABILITY_LABELS.items() if getattr(ChattoAdapter, name, None) is not getattr(BasePlatformAdapter, name, None) ] if ChattoAdapter.supports_code_blocks: found.append("code blocks") if ChattoAdapter.supports_status_text: found.append("custom status text") found.append("presence (refreshed while connected)") return found def register(ctx) -> None: """Plugin entry point — called by the Hermes plugin system.""" logger.info("Registering Chatto platform plugin on Hermes Agent") for capability in _capabilities(): logger.info("Chatto capability: %s", capability) logger.info("ChattoConfiguration.allowed_users.env_name: %s", ChattoConfiguration.allowed_users.env_name) ctx.register_platform( name=ChattoConstants.PLATFORM_NAME, # this will be the config.yaml key. label=ChattoConstants.PLATFORM_LABEL, adapter_factory=hermes_adapter_factory, check_fn=hermes_check_fn, validate_config=hermes_validate_config, is_connected=hermes_is_connected, install_hint=ChattoConstants.INSTALL_HINT, env_enablement_fn=hermes_env_enablement_fn, setup_fn=hermes_setup_fn, cron_deliver_env_var=ChattoConfiguration.home_channel.env_name, standalone_sender_fn=hermes_standalone_sender_fn, allowed_users_env=ChattoConfiguration.allowed_users.env_name, allow_all_env=ChattoConfiguration.allow_all_users.env_name, max_message_length=ChattoConstants.MAX_MESSAGE_LENGTH, emoji="💬", allow_update_command=True, pii_safe=False, platform_hint=( "Using the 'Hermes Chatto Platform Plugin' you connect to a Chatto server. " "Authorized admins and users will contact you and call you his Hermes Agent. They are " "natural persons and thus responsible for what they do in terms of rights. You compute " "on their behalf. They _may_ address you by @-mentioning your name. If configured, " "you also react without a @-mention. Direct messages reach you without a mention." "Keep responses conversational. Markdown is supported. " "Include MEDIA:/absolute/path/to/file in your response to refer to our local files. Images " "(.png, .jpg, .gif, .webp) arrive as inline pictures, videos (.mp4, .mov, .webm) as " "video attachments, audio as a voice bubble, anything else as a downloadable document. " "Do NOT use markdown image syntax for local files. Local files always go through MEDIA:. " "Several images in one response are bundled into a single message. " ), )