Intelligent Data Analysis Assistant

LLM-powered intelligent data analysis assistant supporting natural language queries, SQL generation, visualization, and multi-turn conversation. Suitable for business analysis, report automation, and data exploration. Supports MySQL, PostgreSQL, Snowflake, and Excel/JSON file reading.

Talk to your data through natural language. No SQL or technical background required for data query, analysis, and visualization.

Workflow

text

User question → Parse intent → Generate SQL → Execute query → Analyze results → Visualize → Output conclusions

Core Capabilities

1. Natural Language → SQL

Chinese questions auto-converted to SQL:

User Question	SQL
"Sales by region last month?"	SELECT region, SUM(amount) FROM sales WHERE month='2026-04' GROUP BY region
"Which product has the highest return rate?"	SELECT product, COUNT(*) FROM orders WHERE status='returned' GROUP BY product ORDER BY 2 DESC LIMIT 1
"Compare user growth vs same period last year"	SELECT DATE_TRUNC('month', created_at), COUNT(*) FROM users WHERE created_at >= NOW() - INTERVAL '1 year' GROUP BY 1 ORDER BY 1

2. Data Visualization

Results output in two layers:

• Layer 1: Inline Markdown summary (metrics table + ASCII trend + conclusions)
• Layer 2: Standalone HTML page (Chart.js interactive charts), see references/visualization-template.md

3. Multi-turn Conversation

Mode	Description
Refine	"Only show East China" → append filter
Switch dimension	"Group by month" → re-aggregate
Root cause	"Why did it drop?" → drill down
Compare	"vs last quarter?" → time comparison

4. File Data Reading

Supports Excel (.xlsx/.xls), JSON/JSONL, CSV file reading. See references/data-sources.md.

5. Database Connections

MySQL / PostgreSQL / Snowflake / SQLite / BigQuery / Redshift. See references/data-sources.md.

Output Format

Inline chat output:

text

📊 Results: {title}
─────────────────────────────
{metrics table}

📈 Trend:
{ASCII trend bars}

📋 Analysis:
1. ...

For charts, auto-generate HTML page → write to {domain}_chart.html → report path.

Notes

• SQL limited to read-only SELECT
• Privacy fields auto-masked
• Large datasets prompt for LIMIT
• Vague questions trigger clarifying questions
• Uses mock data when no data source configured
• File reading auto-outputs overview (row count, columns, types, first 5 rows)