Add docs for new non-WS message fetching and remove dead funcs

AGENTS.md

@@ -228,6 +228,7 @@ All endpoints are prefixed with `/api` (e.g., `/api/health`).
 | POST | `/api/packets/maintenance` | Delete old packets (cleanup) |
 | POST | `/api/contacts/{key}/mark-read` | Mark contact conversation as read |
 | POST | `/api/channels/{key}/mark-read` | Mark channel as read |
+| GET | `/api/read-state/unreads` | Server-computed unread counts, mentions, last message times |
 | POST | `/api/read-state/mark-all-read` | Mark all conversations as read |
 | GET | `/api/settings` | Get app settings |
 | PATCH | `/api/settings` | Update app settings |
@@ -266,7 +267,7 @@ Read state (`last_read_at`) is tracked **server-side** for consistency across de
 - Stored as Unix timestamp in `contacts.last_read_at` and `channels.last_read_at`
 - Updated via `POST /api/contacts/{key}/mark-read` and `POST /api/channels/{key}/mark-read`
 - Bulk update via `POST /api/read-state/mark-all-read`
-- Frontend compares `last_read_at` with message `received_at` to count unreads
+- Aggregated counts via `GET /api/read-state/unreads` (server-side computation)
 
 **State Tracking Keys (Frontend)**: Generated by `getStateKey()` for message times (sidebar sorting):
 - Channels: `channel-{channel_key}`

app/AGENTS.md

@@ -35,7 +35,7 @@ app/
     ├── channels.py   # Channel CRUD, radio sync, mark-read
     ├── messages.py   # Message list and send (direct/channel)
     ├── packets.py    # Raw packet endpoints, historical decryption
-    ├── read_state.py # Bulk read state operations (mark-all-read)
+    ├── read_state.py # Read state: unread counts, mark-all-read
     ├── settings.py   # App settings (max_radio_contacts)
     └── ws.py         # WebSocket endpoint at /api/ws
 ```
@@ -104,6 +104,10 @@ await ws_manager.broadcast("message", {"id": 1, "text": "Hello"})
 
 Event types: `health`, `contacts`, `channels`, `message`, `contact`, `raw_packet`, `message_acked`, `error`
 
+**Note:** The WebSocket initial connect only sends `health`. Contacts and channels are fetched
+via REST (`GET /api/contacts`, `GET /api/channels`) for faster parallel loading. The WS still
+broadcasts real-time `contacts`/`channels` updates when data changes.
+
 Helper functions for common broadcasts:
 
 ```python
@@ -490,6 +494,7 @@ All endpoints are prefixed with `/api`.
 - `DELETE /api/channels/{key}` - Delete channel
 
 ### Read State
+- `GET /api/read-state/unreads?name=X` - Server-computed unread counts, mention flags, and last message times
 - `POST /api/read-state/mark-all-read` - Mark all contacts and channels as read
 
 ### Messages

app/repository.py

@@ -497,59 +497,6 @@ class MessageRepository:
             acked=row["acked"],
         )
 
-    @staticmethod
-    async def get_bulk(
-        conversations: list[dict],
-        limit_per_conversation: int = 100,
-    ) -> dict[str, list["Message"]]:
-        """Fetch messages for multiple conversations in one query per conversation.
-
-        Args:
-            conversations: List of {type: 'PRIV'|'CHAN', conversation_key: string}
-            limit_per_conversation: Max messages to return per conversation
-
-        Returns:
-            Dict mapping 'type:conversation_key' to list of messages
-        """
-        result: dict[str, list[Message]] = {}
-
-        for conv in conversations:
-            msg_type = conv.get("type")
-            conv_key = conv.get("conversation_key")
-            if not msg_type or not conv_key:
-                continue
-
-            key = f"{msg_type}:{conv_key}"
-
-            cursor = await db.conn.execute(
-                """
-                SELECT * FROM messages
-                WHERE type = ? AND conversation_key LIKE ?
-                ORDER BY received_at DESC
-                LIMIT ?
-                """,
-                (msg_type, f"{conv_key}%", limit_per_conversation),
-            )
-            rows = await cursor.fetchall()
-            result[key] = [
-                Message(
-                    id=row["id"],
-                    type=row["type"],
-                    conversation_key=row["conversation_key"],
-                    text=row["text"],
-                    sender_timestamp=row["sender_timestamp"],
-                    received_at=row["received_at"],
-                    paths=MessageRepository._parse_paths(row["paths"]),
-                    txt_type=row["txt_type"],
-                    signature=row["signature"],
-                    outgoing=bool(row["outgoing"]),
-                    acked=row["acked"],
-                )
-                for row in rows
-            ]
-
-        return result
-
     @staticmethod
     async def get_unread_counts(name: str | None = None) -> dict:
         """Get unread message counts, mention flags, and last message times for all conversations.

app/routers/messages.py

@@ -38,19 +38,6 @@ async def list_messages(
     )
 
 
-@router.post("/bulk", response_model=dict[str, list[Message]])
-async def get_messages_bulk(
-    conversations: list[dict],
-    limit_per_conversation: int = Query(default=100, ge=1, le=1000),
-) -> dict[str, list[Message]]:
-    """Fetch messages for multiple conversations in one request.
-
-    Body should be a list of {type: 'PRIV'|'CHAN', conversation_key: string}.
-    Returns a dict mapping 'type:conversation_key' to list of messages.
-    """
-    return await MessageRepository.get_bulk(conversations, limit_per_conversation)
-
-
 @router.post("/direct", response_model=Message)
 async def send_direct_message(request: SendDirectMessageRequest) -> Message:
     """Send a direct message to a contact."""

frontend/AGENTS.md

@@ -90,10 +90,13 @@ The `preferences_migrated` flag prevents duplicate migrations.
 
 ### State Flow
 
-1. **WebSocket** pushes real-time updates (health, contacts, channels, messages)
+1. **REST API** fetches initial data on mount in parallel (config, settings, channels, contacts, unreads)
-2. **REST API** fetches initial data and handles user actions
+2. **WebSocket** pushes real-time updates (health, messages, contact changes, raw packets)
 3. **Components** receive state as props, call handlers to trigger changes
 
+**Note:** Contacts and channels are loaded via REST on mount (not from WebSocket initial push).
+The WebSocket only sends health on initial connect, then broadcasts real-time updates.
+
 ### Conversation Header
 
 For contacts, the header shows path information alongside "Last heard":
@@ -229,8 +232,8 @@ interface Message {
 }
 
 interface Conversation {
-  type: 'contact' | 'channel' | 'raw' | 'map';
+  type: 'contact' | 'channel' | 'raw' | 'map' | 'visualizer';
-  id: string;              // PublicKey for contacts, ChannelKey for channels, 'raw'/'map' for special views
+  id: string;              // PublicKey for contacts, ChannelKey for channels, 'raw'/'map'/'visualizer' for special views
   name: string;
 }
 
@@ -397,11 +400,8 @@ for local state tracking, while `conversation_key` is the raw database field.
 Unread tracking uses server-side `last_read_at` timestamps for cross-device consistency:
 
 ```typescript
-// Contacts and channels include last_read_at from server
+// Fetch aggregated unread counts from server (replaces bulk message fetch + client-side counting)
-interface Contact {
+await api.getUnreads(myName);  // Returns { counts, mentions, last_message_times }
-  // ...
-  last_read_at: number | null;  // Unix timestamp when conversation was last read
-}
 
 // Mark as read via API (called automatically when viewing conversation)
 await api.markContactRead(publicKey);
@@ -409,7 +409,9 @@ await api.markChannelRead(channelKey);
 await api.markAllRead();  // Bulk mark all as read
 ```
 
-Unread count = messages where `received_at > last_read_at`.
+The `useUnreadCounts` hook fetches counts from `GET /api/read-state/unreads` on mount and
+when channels/contacts change. Real-time increments are still tracked client-side via WebSocket
+message events. The server computes unread counts using `last_read_at` vs `received_at`.
 
 ## Utility Functions