CCAR-F Prompt Engineering and Structured Output — All Questions With Answers

{
  "error": {
    "type": "invalid_request_error",
    "message": "messages: assistant message must alternate with user message"
  }
}

{
  "model": "claude-3-5-sonnet-20240620",
  "max_tokens": 1024,
  "system": "You are a data assistant. Output JSON.",
  "messages": [
    {"role": "user", "content": "Extract key terms from: [TEXT]"}
  ]
}

{
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "<thought>Identifying user intent: query extraction.</thought>{\"query\": \"weather\"}"
    }
  ]
}

{
  "messages": [
    {
      "role": "user",
      "content": "Extract the name and date from this email: 'Hi John, let's meet on Oct 5th.' Respond only in JSON."
    }
  ],
  "system": "You are a data extractor. Always use the key 'person_name' and 'event_date'."
}

<documents>
<document index="1">
<source>{{SOURCE_1}}</source>
<content>{{CONTENT_1}}</content>
</document>
</documents>

{
  "role": "user",
  "content": "Summarize the following document: [DOC_CONTENT]"
}

{
  "role": "user",
  "content": "Rewrite the following paragraph to be more professional: [TEXT]"
}

{"role": "user", "content": "Extract company name and ticker."}, {"role": "assistant", "content": "{\"company\": \"Anthropic\", \"ticker\": \"N/A\"}"}

Error: Output truncated due to token limit. Current schema violation: missing closing brace.