08. Output Formatting

flowchart TD
    subgraph UNFORMATTED["Without Format Instruction"]
        U1["User: Compare React and Vue"] --> U2["❌ Random format\nParagraph? List? Table? Essay?"]
    end

    subgraph FORMATTED["With Format Instruction"]
        F1["User: Compare React and Vue\nFormat: Markdown table"] --> F2["✅ Predictable format\n| Aspect | React | Vue |"]
    end

    style UNFORMATTED fill:#ef4444,color:#fff
    style FORMATTED fill:#22c55e,color:#fff

flowchart LR
    FORMATS["Output Formats"] --> MARKDOWN["Markdown\nDocs, READMEs\nCode blocks, tables"]
    FORMATS --> JSON["JSON\nAPIs, data processing\nProgrammatic parsing"]
    FORMATS --> HTML["HTML\nWeb content\nEmails, reports"]
    FORMATS --> XML["XML\nLegacy systems\nComplex hierarchies"]
    FORMATS --> CSV["CSV\nSpreadsheets\nData analysis"]
    FORMATS --> YAML["YAML\nConfig files\nFrontmatter"]
    FORMATS --> CODE["Code\nDirect output\nNo wrapping"]

    style FORMATS fill:#8b5cf6,color:#fff
    style MARKDOWN fill:#3b82f6,color:#fff
    style JSON fill:#22c55e,color:#fff
    style HTML fill:#f59e0b,color:#fff
    style XML fill:#ec4899,color:#fff
    style CSV fill:#14b8a6,color:#fff
    style YAML fill:#f97316,color:#fff
    style CODE fill:#ef4444,color:#fff

Format the response as a markdown document with:
- A level-2 heading for each section
- Code blocks with language tags
- A table at the end for summary

Explain the difference between let, const, and var in JavaScript.
Format your response as a markdown table with columns:
| Feature | let | const | var |
|---------|-----|-------|-----|
| Scope | ... | ... | ... |
| Reassignable | ... | ... | ... |
| Hoisted | ... | ... | ... |

Return the response as a JSON object with the following structure:
{
  "summary": "brief overview",
  "key_points": ["point1", "point2"],
  "recommendation": "your recommendation"
}

Extract the following information from this invoice and return it as JSON:
{
  "vendor_name": "string",
  "invoice_date": "YYYY-MM-DD",
  "total_amount": "number",
  "line_items": [
    {
      "description": "string",
      "quantity": "number",
      "unit_price": "number"
    }
  ]
}

Invoice text:
[invoice text here]

{
  "vendor_name": "Acme Corp",
  "invoice_date": "2024-01-15",
  "total_amount": 1500.00,
  "line_items": [
    {
      "description": "Web development services - January",
      "quantity": 40,
      "unit_price": 37.50
    }
  ]
}

Return the response as HTML. Use semantic tags and inline styles.

Create a pricing card for a SaaS product with three tiers: Basic ($9/mo),
Pro ($29/mo), and Enterprise ($99/mo). Return as HTML with inline styles
for a clean, modern look.

Return the configuration as XML:
<config>
  <database>
    <host>localhost</host>
    <port>5432</port>
  </database>
  <features>
    <feature enabled="true">authentication</feature>
  </features>
</config>

Return the data as CSV with headers:
name,email,role,department,start_date

Format the API documentation as YAML:
endpoints:
  - path: /users
    method: GET
    description: List all users

flowchart TD
    Q1["Who consumes the output?"]
    Q1 -->|"Humans reading"| Q2["How complex?"]
    Q1 -->|"Machines parsing"| JSON["JSON or YAML"]

    Q2 -->|"Simple"| MARKDOWN["Markdown"]
    Q2 -->|"Rich content"| HTML["HTML"]
    Q2 -->|"Code"| CODE["Code blocks"]
    Q2 -->|"Data table"| TABLE["Table or CSV"]

❌ "Return a JSON object with the user's name, email, and role."

✅ "Return as JSON:
{
  "name": "string",
  "email": "email format",
  "role": "admin | user | viewer"
}"

✅ "Format each item as: [Name] — [Role] — [Years of Experience]
   Example: Jane Smith — Senior Engineer — 8 years"

✅ "Return ONLY the JSON object. No markdown formatting, no code blocks,
    no explanation text before or after the JSON."

✅ "Return a 3-sentence summary in a single paragraph."

✅ "Remember: Return ONLY JSON. No other text."

❌ "Extract the key data from this email."

✅ "Extract the following from this email and return as JSON:
{
  "sender": "sender's email address",
  "subject": "email subject line",
  "urgency": "high | medium | low",
  "action_items": ["item1", "item2"],
  "deadline": "YYYY-MM-DD or null if none"
}
Return ONLY the JSON object."

❌ "Explain this API endpoint."

✅ "Document this API endpoint in markdown with the following sections:
## Endpoint
## Method
## Request Body
## Response
## Example
## Error Codes

Use a code block for the example request/response."

{
  "model": "gpt-4o",
  "response_format": { "type": "json_object" },
  "messages": [
    {"role": "system", "content": "You are a helpful assistant that outputs JSON."},
    {"role": "user", "content": "Extract the name and age from: John is 30 years old"}
  ]
}