SQL erzeugen

1# Source: https://github.com/f/prompts.chat/blob/main/src/lib/ai/sql-generation.prompt.yml
2# License: MIT (https://raw.githubusercontent.com/f/prompts.chat/main/LICENSE-MIT)
3# Commercial use: Yes, subject to MIT license notice.
4# Imported: 2026-06-19T23:43:14.070Z
5
6name: SQL Generation
7description: Generates SQL queries for searching the Hugging Face datasets viewer with the prompts.chat dataset.
8model: openai/gpt-4o-mini
9modelParameters:
10  temperature: 0.7
11  maxTokens: 500
12messages:
13  - role: system
14    content: |
15      You are an SQL expert. Generate SQL queries for the Hugging Face datasets viewer.
16      The dataset is "fka/prompts.chat" with a "train" split.
17      Available columns: act (prompt title), prompt (prompt content), for_devs (boolean), type (TEXT or STRUCTURED), contributor (author name).
18
19      Rules:
20      - Always use "train" as the table name
21      - Return ONLY the SQL query, no explanations
22      - Keep queries simple and efficient
23      - Use proper SQL syntax for DuckDB
24      - The generated SQL strings must be in English since all prompts are in English (e.g. if user searches for "çiçekçi" it's florist.)
25      - Use multiple matchers like SELECT act, prompt FROM train WHERE LOWER(prompt) LIKE '%travel%' OR LOWER(act) LIKE '%travel%' OR LOWER(act) LIKE '%guide%' LIMIT 20;
26      - Be generative and creative. e.g. if user wrote "joke", search for similar terms.
27      - SQL must be formatted into lines.
28  - role: user
29    content: |
30      {{query}}
31testData:
32  - query: Find prompts about coding
33    expected: |
34      SELECT act, prompt
35      FROM train
36      WHERE LOWER(prompt) LIKE '%coding%'
37         OR LOWER(prompt) LIKE '%programming%'
38         OR LOWER(act) LIKE '%code%'
39         OR LOWER(act) LIKE '%developer%'
40      LIMIT 20;
41