Tabellen und strukturierte Daten anfordern

Tabellen als Ausgabeformat

Grundlegende Tabellenanfrage

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'Compare Python, JavaScript, and Go for building REST APIs. '
            'Format as a markdown table with columns: '
            'Language, Performance, Learning Curve, Ecosystem, Best Use Case. '
            'One row per language. Be concise — max 8 words per cell.'
        )
    }]
)
print(response.choices[0].message.content)

Vergleichstabellen

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=400,
    messages=[{
        'role': 'user',
        'content': (
            'Create a comparison table for three cloud database options: '
            'Supabase, PlanetScale, and Neon.\n'
            'Rows: one per database.\n'
            'Columns: Free tier storage, Scaling model, Pricing at 10GB, '
            'Branching support, SQL compatibility.\n'
            'Format: markdown table. Cell content: max 6 words each.'
        )
    }]
)
print(response.content[0].text)

Feature-Matrix-Tabellen

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'Create a feature matrix table for our SaaS pricing tiers: Free, Pro, Enterprise.\n'
            'Rows: API access, Custom domains, SSO login, SLA guarantee, Priority support, '
            'Data export, Team collaboration, Audit logs.\n'
            'Columns: Free, Pro, Enterprise.\n'
            'Use: YES / NO / PARTIAL for each cell.\n'
            'Format: markdown table.'
        )
    }]
)
print(response.choices[0].message.content)

Datentabellen aus Text

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

meeting_notes = (
    'Sarah will handle the API documentation by next Friday. '
    'John needs to fix the authentication bug — aim for Wednesday. '
    'Maria is responsible for the mobile UI update, due in 2 weeks. '
    'The DevOps team (lead: Alex) must set up staging by Monday. '
    'Carlos will write unit tests — deadline: end of sprint (Thursday).'
)

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=300,
    messages=[{
        'role': 'user',
        'content': (
            'Extract all action items from the meeting notes below. '
            'Format as a markdown table with columns: Task, Owner, Deadline.\n\n'
            f'Meeting notes:\n{meeting_notes}'
        )
    }]
)
print(response.content[0].text)

CSV-Ausgabe

import openai
import csv
import io

client = openai.OpenAI(api_key='sk-your-key-here')

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'Generate 5 sample customer support tickets for an e-commerce store. '
            'Output as CSV with header row. '
            'Columns: ticket_id, customer_name, issue_type, priority, status. '
            'issue_type values: SHIPPING / BILLING / PRODUCT / RETURNS. '
            'priority values: LOW / MEDIUM / HIGH. '
            'status values: OPEN / IN_PROGRESS / RESOLVED. '
            'No markdown formatting — raw CSV only.'
        )
    }]
)

csv_text = response.choices[0].message.content.strip()
reader = csv.DictReader(io.StringIO(csv_text))
for row in reader:
    print(row)

Strukturierte JSON-Ausgabe

import anthropic
import json

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=400,
    messages=[{
        'role': 'user',
        'content': (
            'Generate 3 fictional user profiles for testing a fitness app. '
            'Output as a JSON array. '
            'Each object must have exactly these keys: '
            'id (integer), name (string), age (integer 18-65), '
            'fitness_level ("BEGINNER" | "INTERMEDIATE" | "ADVANCED"), '
            'goals (array of strings, max 3), weekly_sessions (integer 1-7). '
            'Output raw JSON only — no explanation, no markdown code fences.'
        )
    }]
)

try:
    data = json.loads(response.content[0].text)
    print(f'Parsed {len(data)} profiles successfully.')
    for profile in data:
        print(f'  {profile["name"]}, {profile["age"]}, {profile["fitness_level"]}')
except json.JSONDecodeError as e:
    print('JSON parse error:', e)

Regeln für den Zellinhalt festlegen

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'Rate the following programming languages for these criteria.\n'
            'Languages (rows): Python, Rust, TypeScript, Go.\n'
            'Criteria (columns): Speed, Safety, Developer Experience, Job Market, Learning Curve.\n'
            'Cell format: integer score 1-10 only. No words, no explanation inside cells.\n'
            'After the table, add one sentence explaining your scoring philosophy.'
        )
    }]
)
print(response.choices[0].message.content)

Transponierte Tabellen

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=300,
    messages=[{
        'role': 'user',
        'content': (
            'Compare two job candidates: Alice and Bob. '
            'Create a transposed table where candidates are COLUMNS and attributes are ROWS. '
            'Attributes (rows): Years Experience, Primary Language, Education, '
            'Previous Employer, Salary Expectation Range. '
            'Format: markdown table. Cell content: factual, max 6 words.\n\n'
            'Alice: 7 years Python/ML, MS Computer Science MIT, ex-Google, expects $180-220k.\n'
            'Bob: 4 years JavaScript/Node, BS from UC Berkeley, ex-Stripe startup, expects $140-170k.'
        )
    }]
)
print(response.content[0].text)

Wann Sie Tabellen NICHT verwenden sollten

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

# Table works: structured comparison
table_prompt = (
    'Compare MySQL vs PostgreSQL. '
    'Format: markdown table, columns: Feature, MySQL, PostgreSQL. '
    '6 rows covering: JSON support, Full-text search, Replication, Transactions, License, Best for.'
)

# No table needed: simple 2-option choice with prose reasoning
not_table_prompt = (
    'Should I use SQLite or PostgreSQL for a personal hobby project '
    'that will have at most 5 users? 2-sentence answer, no table.'
)

for label, prompt in [('TABLE APPROPRIATE', table_prompt), ('TABLE OVERKILL', not_table_prompt)]:
    response = client.chat.completions.create(
        model='gpt-4o', max_tokens=200,
        messages=[{'role': 'user', 'content': prompt}]
    )
    print(f'--- {label} ---')
    print(response.choices[0].message.content.strip()[:300])
    print()

Tabellen mit anderen Formaten kombinieren

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=500,
    messages=[{
        'role': 'user',
        'content': (
            'Compare three Python testing frameworks: pytest, unittest, and nose2. '
            'Structure your response as:\n'
            '1. One-sentence intro explaining the comparison scope.\n'
            '2. Markdown table: columns = Framework, Ease of Use, Plugin Ecosystem, '
            'Speed, Community Activity. Rows = one per framework.\n'
            '3. One-sentence recommendation for a team starting a new project.'
        )
    }]
)
print(response.content[0].text)

Wissenscheck

Tabellen und strukturierte Daten – Zusammenfassung