analysis

name	analysis
display_name	AI 분석
description	Manually trigger AI analysis — conversation insights, spending patterns, memory compaction. Runs automatically on schedule; only invoke when user explicitly requests an analysis or memory update.
version	2.0.0
emoji	📊
category	productivity
enabled_by_default	true
requires_api_key	false
platforms	web, telegram, api
uses_provider	false
allowed-tools	["Bash","exec"]
triggers	{"keywords":["분석해줘","인사이트","패턴","요약해줘","메모리 업데이트","대화 분석","지출 패턴","통계","리포트","analysis","insight","pattern","summarize","memory update","weekly report"],"when_to_use":["User explicitly requests an AI analysis or insights report","User asks to update or compact memory","User asks for spending pattern analysis or conversation summary","User says \"분석해줘\" or \"인사이트 보여줘\""],"not_for":["Automatic scheduled analysis (runs on its own cron schedule)","Real-time data queries (use finance or memory skill instead)"]}

AI Analysis (Manual Trigger)

Analysis runs automatically on schedule. Only trigger manually when the user explicitly asks.

User intent	Type	Schedule (auto)
User wants analysis of today's conversations or daily insights	conversation	Daily 23:00
User wants spending pattern analysis or expense insights	patterns	Daily 06:00
User wants memory compaction or old diary cleanup	compact	Weekly Mon 05:00

Manual trigger via agent API

The scheduler exposes triggerAnalysis(type, userId) — but this is a TypeScript function, not a CLI script. To manually trigger from conversation, use the gateway admin endpoint or restart the agent service (scheduler runs automatically on bind).

Note: Results appear in the AI 기억 menu after the job completes. Data is stored in knowledge_base table with keys:

• conversation:analysis:YYYY-MM-DD
• pattern:analysis_result
• memory:weekly_summary:YYYY-Www