Skip conv_apps suggestion when user has preferred app

[beastoin]

Parent: #4635

Problem

_trigger_apps() always calls get_suggested_apps_for_conversation() (LLM call) even when the user already has a preferred_app_id set in Redis. The suggestion result is stored but never used.

File: backend/utils/conversations/process_conversation.py:300-326

# Always runs suggestion first
if not conversation.suggested_summarization_apps:
    with track_usage(uid, Features.CONVERSATION_APPS):
        suggested_apps, reasoning = get_suggested_apps_for_conversation(...)

# Then checks preferred app anyway
preferred_app_id = redis_db.get_user_preferred_app(uid)
if preferred_app_id and preferred_app_id in all_apps_dict:
    app_to_run = all_apps_dict.get(preferred_app_id)  # Uses this, ignoring suggestion

Proposed Change

Check for preferred app FIRST. If user has one, skip the suggestion LLM call entirely.

# Check preferred app first
preferred_app_id = redis_db.get_user_preferred_app(uid)
if preferred_app_id and preferred_app_id in all_apps_dict:
    app_to_run = all_apps_dict[preferred_app_id]
else:
    # Only suggest if no preferred app
    if not conversation.suggested_summarization_apps:
        suggested_apps, reasoning = get_suggested_apps_for_conversation(...)

Impact

• Eliminates 1 LLM call per conversation for users with preferred apps
• Estimated 2-3% reduction in total costs
• Uses llm_mini so cost per call is small, but volume is high

Risk

Low — suggested_summarization_apps field won't be populated for these users, but it's unused when preferred app is set.

[beastoin]

📝 The proposed solutions here are based on code analysis and have not been verified by the CTO. If you have a better approach to address this, feel free to propose it.

[beastoin]

Updated: Feb 5 BQ Data

conv_apps accounts for 9.8% of calls and 13.5% of total cost. The suggestion call (gpt-4.1-mini) is cheap per-call but runs on every conversation — skipping it when a preferred app exists would reduce call volume with zero quality impact.