duckdb/SKILL.md

name, description, metadata

name	description	metadata
duckdb	DuckDB embedded analytical database with HTTP API	version vibestack 1.0.0 main false

DuckDB Skill

DuckDB - fast in-process analytical database with a simple HTTP API.

Features

• Embedded OLAP database (no separate server process)
• Query CSV, Parquet, JSON files directly
• SQL interface via HTTP API
• Persistent storage option
• Auto-registers with Caddy if present

Configuration

Variable	Description	Default
DUCKDB_PORT	HTTP API port	8432
DUCKDB_DATABASE	Database file path	:memory:
DUCKDB_DATA_DIR	Directory for data files	/data/duckdb
DUCKDB_DOMAIN	Domain for Caddy auto-config	(none)
DUCKDB_READ_ONLY	Read-only mode	false

HTTP API

Execute Query

# Simple query
curl -X POST http://localhost:8432/query \
  -H "Content-Type: application/json" \
  -d '{"sql": "SELECT 1 + 1 AS result"}'

# Query CSV file
curl -X POST http://localhost:8432/query \
  -d '{"sql": "SELECT * FROM read_csv_auto(\"/data/duckdb/sales.csv\") LIMIT 10"}'

# Query Parquet file
curl -X POST http://localhost:8432/query \
  -d '{"sql": "SELECT * FROM read_parquet(\"/data/duckdb/events.parquet\")"}'

Response Format

{
  "success": true,
  "columns": ["result"],
  "rows": [[2]],
  "row_count": 1,
  "time_ms": 0.5
}

Use Cases

Analytics on Log Data

-- Query JSON logs
SELECT
  json_extract_string(line, '$.level') as level,
  count(*) as count
FROM read_json_auto('/var/log/supervisor/*.log')
GROUP BY level;

Query Remote Data

-- Query remote Parquet (S3, HTTP)
SELECT * FROM read_parquet('https://example.com/data.parquet');

-- Query remote CSV
SELECT * FROM read_csv_auto('https://example.com/data.csv');

Create Persistent Tables

-- Create table
CREATE TABLE events AS
SELECT * FROM read_parquet('/data/duckdb/events.parquet');

-- Query table
SELECT date_trunc('hour', timestamp) as hour, count(*)
FROM events
GROUP BY 1 ORDER BY 1;

CLI Access

# Interactive shell
duckdb /data/duckdb/analytics.db

# One-off query
duckdb -c "SELECT * FROM 'data.csv' LIMIT 5"

Extensions

DuckDB supports extensions for additional functionality:

-- Install and load extensions
INSTALL httpfs;
LOAD httpfs;

-- Now query S3/HTTP directly
SELECT * FROM read_parquet('s3://bucket/data.parquet');