[gvieralopez]

Makes the deployed bot work without Blizzard credentials and within 256 MB limit, plus a few features found while testing. One commit per change:

1. Fetch profiles and ladders from SC2 Pulse.
   • The daily fetch and /fetch asked the Blizzard API, which needs credentials, and its public fallback is gone.
   • SC2 Pulse gives every stored profile's current-season 1v1 ladders in one call per season.
   • A ladder is stored only when its MMR or game count changed, dated by when it was last played.
   • Every fetch also searches each user's Battle Tag again, so new profiles appear by themselves.
   • /link, the Blizzard client and the BLIZZARD_CLIENT_ID, BLIZZARD_CLIENT_SECRET and REALM_ID settings are removed.
   • China gets a flag and a chart style.
2. Make the replies shorter and consistent.
   • Every command is a tappable /command.
   • Monospace is kept for what is worth copying.
   • Usernames get their @, and users without a username are stored as a link again.
   • /help is grouped into You and Everyone.
3. Send long replies in several messages. Replies are cut between lines to fit Telegram's limits: 4096 characters per message, 1024 per photo caption. The daily post pins its first part.
4. Add /ping to mention the group.
   • In a group, it mentions the bot's users who are members of that group, except the caller.
   • It sends four mentions per message, because Telegram only notifies the first four of a message.
   • In a private chat it answers with its help.
5. Read tapped help links in groups. In a group, a tapped link arrives as /help_fetch@bot, which was read as help for a command called fetch@bot.
6. Replace aiogram with python-telegram-bot.
   • aiogram builds all ~450 of its models at import, about 140 MB. python-telegram-bot costs about 20 MB.
   • Inside the image, the bot starts at 119 MB instead of 247 MB.
   • After every user's /profile it peaks at 145 MB instead of 272 MB.
   • The charts are byte-for-byte the same.
   • The daily fetch runs on python-telegram-bot's job queue, in the server's time zone.
7. Prepare version 0.2.0. Removing a command and three settings makes the next release 0.2.0.

Testing

• make qa passes (108 tests, 98% coverage), including from a fresh clone.
• The daily fetch was run against a copy of the production database.
• The image was run as a test bot under --memory 256m --read-only and tried in Telegram.

After merge

• Redeploy 0.2.0 on the homelab.
• Drop the Blizzard variables from the stack.
• Set LOG_CHANNEL_ID to the group.

🤖 Generated with Claude Code