@LazuliKao — need a check on the production instance. This started as a question about the Telegram rich-message change (#20, #21), but the timeline rules that out.
What the NBTCA Telegram group shows:
| Topic |
Last message from NBoT |
| Repair |
30 March, an issue card |
| 开发日志 |
8 May, a calendar reminder; the last GitHub push card is earlier, around 4 May |
| Minecraft |
28 April, a server join/leave line |
| General |
still current today — but those are from GitHub's official Telegram integration (@githubbot), not from us |
So NBoT has posted nothing for about four months, well before the rich-message work landed on 5 September. #20 and #21 are not the cause, and deploying them will not fix it.
It is also not only the GitHub bridge: the Calendar reminders and the Minecraft broadcasts stopped in the same window. Three unrelated plugins going quiet together points at the process or its connection, not at one plugin.
Questions:
- Is the production instance still running, and since when?
- If it is running, is
ServerlessMQ connected? It logs 连接成功 / 重新连接成功 on connect and 连接断开 on drop (src/HuaJiBot.NET.Plugin.GitHubBridge/PluginMain.cs:78-89). No webhook reaches the bridge without it.
- Did anything change in early May — host, bot token, the relay at
Config.Address, or the bot's permissions in the group?
- Should @githubbot in General stay? It already covers GitHub events, which changes how urgent this is.
One thing to fix before the next deploy, once the bot is back: SendRichMessageAsync logs and rethrows on failure instead of falling back to the image card (src/HuaJiBot.NET.Adapter.Telegram/TelegramAdapter.cs:226), so a rejected sendRichMessage posts nothing at all. Left as is, one bad Bot API call would look exactly like the silence we are debugging now.
@LazuliKao — need a check on the production instance. This started as a question about the Telegram rich-message change (#20, #21), but the timeline rules that out.
What the NBTCA Telegram group shows:
So NBoT has posted nothing for about four months, well before the rich-message work landed on 5 September. #20 and #21 are not the cause, and deploying them will not fix it.
It is also not only the GitHub bridge: the Calendar reminders and the Minecraft broadcasts stopped in the same window. Three unrelated plugins going quiet together points at the process or its connection, not at one plugin.
Questions:
ServerlessMQconnected? It logs 连接成功 / 重新连接成功 on connect and 连接断开 on drop (src/HuaJiBot.NET.Plugin.GitHubBridge/PluginMain.cs:78-89). No webhook reaches the bridge without it.Config.Address, or the bot's permissions in the group?One thing to fix before the next deploy, once the bot is back:
SendRichMessageAsynclogs and rethrows on failure instead of falling back to the image card (src/HuaJiBot.NET.Adapter.Telegram/TelegramAdapter.cs:226), so a rejectedsendRichMessageposts nothing at all. Left as is, one bad Bot API call would look exactly like the silence we are debugging now.