chore: turn off tweet aggregation

[dewabisma]

Note

Medium Risk
Operational change: background tweet sync and its alerts stop running, which may break tweet-dependent features even though the HTTP API still starts.

Overview
Disables the background tweet sync loop by removing it from the process’s tokio::select! in main.rs. On startup, TaskMaster now only waits on the HTTP server task instead of racing the server against the tweet synchronizer.

The periodic tweet aggregation path (spawn_tweet_synchronizer) is no longer started, so in-process Twitter pull/sync, related DB updates, and downstream alerting driven by that worker will not run while the API keeps serving.

^{Reviewed by Cursor Bugbot for commit a6dd967. Configure here.}

[n13]

Review

The intent (turning off tweet aggregation) is fine, but the change as-is breaks CI and leaves half-removed code behind.

Blocking

1. CI fails on clippy (failed run) — main is green, so this is introduced by the PR. CI runs cargo clippy --all-targets --all-features -- -D warnings, and removing only the select! arm makes the whole tweet-sync call graph dead code from the binary's perspective:

• unused variable: tweet_synchronizer in main.rs — the service is still constructed but never used
• Dead-code errors cascade from there: get_tweet_sync_interval, AppError::Telegram, extract_sdk_error_type, track_twitter_api_call, track_tweets_pulled, NewTweetPayload, TweetPullUsage, repository methods (get_newest_tweet_id, upsert_many, get_whitelist), and more

2. Decide whether this is a removal or a toggle, and finish accordingly:

• If it's a permanent removal: also delete the now-dead construction in main.rs (tweet_synchronizer, telegram_service, and alert_service which is only consumed by the synchronizer) and the dead code clippy flags.
• If it's a temporary operational off-switch: prefer a config flag (e.g. tweet_sync.enabled) that skips spawning the synchronizer. That keeps the code alive (no dead-code cascade) and lets ops re-enable it without a code change.

Minor

• With one branch left, tokio::select! is unnecessary — server_task.await?? does the same thing.

Non-blocking observation

The /relevant-tweets and /tweet-authors endpoints stay live but will serve stale data, and the Telegram usage alerts driven by the sync loop stop firing. Presumably intended, just flagging it.

Verdict

Request changes — CI must pass; pick removal-or-toggle and complete it.

[n13]

Re-review

The previous blocking issues are resolved — the author went with the permanent-removal route and completed it. CI is now green (formatting, clippy with -D warnings --all-targets --all-features, migrations, and tests all pass; run).

What was removed (clean)

• Services: tweet_synchronizer_service, alert_service, telegram_service (deleted entirely)
• Models/repos: tweet_pull_usage model + repository, TweetPullUsage field on DbPersistence
• Config: TweetSyncConfig, TelegramBotConfig, AlertConfig structs + their [tweet_sync]/[tg_bot]/[alert] sections in all three toml files, and get_tweet_sync_interval
• Errors: AppError::Telegram variant + its IntoResponse arm
• Metrics: track_twitter_api_call, track_tweets_pulled, extract_sdk_error_type, and the twitter metric counters/registrations
• Repo methods only used by the sync loop: get_whitelist, get_newest_tweet_id
• utils/x_url::build_x_status_url

Nice touches

• Correctly gated test-only code behind #[cfg(test)] instead of deleting it: NewTweetPayload, RaidQuestRepository::{find_active, create_select_base_query}, and the upsert_many methods on the tweet/author repos. Keeps the integration tests working without leaving dead production code.
• main.rs now passes None for the rusx api key. Verified safe — the HTTP server's only gateway use is users().get_by_username(...) (OAuth via x_oauth), and test_app_state already passed None, so this is consistent.
• test_db.rs TRUNCATE list updated to drop tweet_pull_usage. Matches the removal.
• Grep across src/ confirms zero dangling references to any removed symbol.

Non-blocking notes (optional)

1. Orphan migration — migrations/009_tweet_pull_usage_table.sql still creates the tweet_pull_usage table on every fresh DB. Leaving applied migrations alone is the right default, but the table is now dead schema. If you want a clean DB, add a follow-up migration (e.g. 011_drop_tweet_pull_usage_table.sql) to drop it; otherwise harmless clutter.
2. PR title — "turn off tweet aggregation" undersells it; this is now a full removal. Consider retitling to "chore: remove tweet aggregation" so history is accurate.
3. /relevant-tweets and /tweet-authors endpoints stay live (the latter still calls the Twitter users API on POST to look up authors). Only the background pull, Telegram raid notifications, and usage-limit alerts are gone — presumably the intent.

Verdict

Approve. Removal is clean and complete, CI passes, no dangling references.