Formatação markdown nas solicitações

Markdown na saída da IA

Solicitando cabeçalhos

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': (
            'Write a technical guide outline for "Getting Started with FastAPI". '
            'Structure: one # H1 title at the top, then 4 ## H2 section headers, '
            'each with 2 ### H3 subsection headers beneath it. '
            'Add one sentence of placeholder content under each H3.'
        )
    }]
)
print(response.content[0].text)

Ênfase em negrito e itálico

import openai

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

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'Explain the concept of idempotency in REST APIs. '
            'Rules:\n'
            '- Bold every technical term on its first occurrence only\n'
            '- Italicize all HTTP method names (GET, POST, PUT, DELETE, PATCH)\n'
            '- 150 words max, flowing prose — no bullets or headers'
        )
    }]
)
print(response.choices[0].message.content)

Blocos de código

```python
print('hello')
```

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': (
            'Show me how to connect to PostgreSQL from Python using psycopg3.\n'
            'Structure:\n'
            '1. Install command in a bash code block.\n'
            '2. Connection example in a python code block with type hints.\n'
            '3. A sample SELECT query in a python code block.\n'
            'Keep each code block under 10 lines. Brief one-sentence intro before each block.'
        )
    }]
)
print(response.content[0].text)

Código em linha

import openai

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

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'Explain the difference between Python list .append() and .extend(). '
            'Rules:\n'
            '- Use inline code for all method names, parameter names, and variable examples\n'
            '- Use a python code block for each demonstration example\n'
            '- Prose sections: max 2 sentences\n'
            '- Do NOT use headers or bullets — flowing prose with code blocks only'
        )
    }]
)
print(response.choices[0].message.content)

Citações em bloco

> This is a blockquote.

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': (
            'Write a security guide section about SQL injection prevention. '
            'Structure:\n'
            '- 2-sentence explanation of the risk\n'
            '- One blockquote containing a real example of vulnerable code (as a note/warning)\n'
            '- 3 bullet points on how to prevent it\n'
            '- One blockquote containing the safe alternative code pattern'
        )
    }]
)
print(response.content[0].text)

Listas aninhadas em marcação

- Main item
  - Sub-item
  - Sub-item
    - Sub-sub-item

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 2-level nested markdown list of AWS services for a web startup. '
            'Level 1: 4 service categories (Compute, Storage, Database, Networking). '
            'Level 2: 3 specific services under each category with a 5-word description. '
            'Format: markdown nested bullets with proper indentation.'
        )
    }]
)
print(response.choices[0].message.content)

Links e imagens

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': (
            'Write a README section for a Python open-source project called "sqlens". '
            'Include:\n'
            '- An image placeholder for a demo screenshot: ![Demo screenshot](docs/demo.png)\n'
            '- At least 2 markdown links: one to the PyPI page, one to the documentation\n'
            '- A badge placeholder using an image link\n'
            '- 3 bullet points of key features\n'
            'Use realistic placeholder URLs (pypi.org/project/sqlens etc).'
        )
    }]
)
print(response.content[0].text)

Linhas horizontais e separadores

import openai

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

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'Write a mini technical specification document for a user authentication API. '
            'Include exactly 3 sections: Overview, Endpoints, Security Requirements. '
            'Separate each section with a --- horizontal rule. '
            'Each section: ## H2 header + 3-5 bullet points of content. '
            'Under Endpoints: use inline code for all route paths and HTTP methods.'
        )
    }]
)
print(response.choices[0].message.content)

Quando a marcação não é renderizada

import anthropic

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

# Check if environment renders markdown before requesting it
rendering_environments = {
    'GitHub':     True,
    'Notion':     True,
    'Obsidian':   True,
    'VS Code':    True,
    'Gmail body': False,  # some markdown, not all
    'Outlook':    False,
    'SMS':        False,
    'Plain text file': False,
}

print('Markdown rendering support:')
for env, renders in rendering_environments.items():
    status = 'RENDERS' if renders else 'DOES NOT RENDER'
    print(f'  {env:<20} {status}')

# Decision: use markdown only when you know it renders
use_markdown = True  # set based on your environment

format_instruction = (
    'Use markdown headers, bold, and code blocks.' if use_markdown
    else 'Plain text only — no markdown symbols.'
)
print('\nFormat instruction:', format_instruction)

Combinando elementos de marcação

import openai

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

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'Write a mini developer guide for the requests Python library. '
            'Use all of the following markdown elements:\n'
            '- # H1 title at the top\n'
            '- ## H2 sections: Installation, Basic Usage, Error Handling\n'
            '- Bold all key terms on first use\n'
            '- Code blocks with python/bash language hints\n'
            '- Inline code for all function names\n'
            '- One blockquote warning about timeout best practice\n'
            '- --- between each section\n'
            'Max 300 words total.'
        )
    }]
)
print(response.choices[0].message.content)

Verificação de conhecimento

Marcação em instruções — recapitulação