json-result-formatting

name	json-result-formatting
description	Format query results to JSON with proper structure and token tracking

JSON Result Formatting

Overview

Formatting retrieved data into standardized JSON output with answer lists and token tracking.

Use Cases

• Writing query results to JSON files
• Formatting answers as lists regardless of count
• Tracking API token consumption
• Creating consistent output files for downstream processing

Required Output Format

{
    "q1": {"answer": ["xxx"], "tokens": 123},
    "q2": {"answer": ["xxx", "yyy"], "tokens": 456},
    "q3": {"answer": [], "tokens": 789}
}

Code Examples

Initialize Result Container

import json

result = {
    "q1": {"answer": [], "tokens": 0},
    "q2": {"answer": [], "tokens": 0},
    "q3": {"answer": [], "tokens": 0}
}

Add Single Answer

def add_answer(result, question_key, answer_items, tokens=0):
    """Add answer as list (always list format)"""
    # Ensure answer_items is a list
    if isinstance(answer_items, str):
        answer_items = [answer_items]
    elif not isinstance(answer_items, list):
        answer_items = list(answer_items)

    result[question_key] = {
        "answer": answer_items,
        "tokens": int(tokens)
    }
    return result

# Usage
result = add_answer(result, "q1", ["eid_1e9356f5"], tokens=150)
result = add_answer(result, "q2", employee_ids_list, tokens=200)

Write to JSON File

import json

def write_result_file(result, filepath):
    """Write result to JSON file"""
    with open(filepath, 'w') as f:
        json.dump(result, f, indent=4)
    print(f"Results written to {filepath}")

# Usage
write_result_file(result, '/root/answer.json')

Track Token Usage (Without API)

import json

# For local data processing, estimate tokens
def estimate_tokens_from_text(text):
    """Rough estimation: ~4 characters per token"""
    return len(text) // 4

# Better: track actual API usage
def track_tokens(usage_dict):
    """Track from API response"""
    if hasattr(usage_dict, 'input_tokens'):
        return usage_dict.input_tokens + usage_dict.output_tokens
    return 0

Complete Result Building Flow

import json

# Initialize
result = {
    "q1": {"answer": [], "tokens": 0},
    "q2": {"answer": [], "tokens": 0},
    "q3": {"answer": [], "tokens": 0}
}

# Q1: Authors and reviewers
authors = ["eid_1e9356f5"]
reviewers = ["eid_06cddbb3", "eid_99835861"]
result["q1"]["answer"] = authors + reviewers
result["q1"]["tokens"] = 150  # Estimated or tracked

# Q2: Competitor insights team members
competitor_team = ["eid_xxx", "eid_yyy"]
result["q2"]["answer"] = competitor_team
result["q2"]["tokens"] = 200

# Q3: Demo URLs
urls = ["https://example.com/demo1", "https://example.com/demo2"]
result["q3"]["answer"] = urls
result["q3"]["tokens"] = 100

# Write output
with open('/root/answer.json', 'w') as f:
    json.dump(result, f, indent=4)

Best Practices

1. Always Use Lists

   • Single item: ["value"] not "value"
   • Multiple items: ["item1", "item2"]
   • Empty: [] not null

2. Token Tracking

   • For Claude API: Use response.usage.input_tokens + response.usage.output_tokens
   • For local processing: Track API calls made and sum their tokens
   • If no API: Use 0 or estimate conservatively

3. Data Validation

   # Ensure no duplicates
   result["q1"]["answer"] = list(set(result["q1"]["answer"]))
   
   # Ensure proper types
   result["q2"]["tokens"] = int(result["q2"]["tokens"])
   

4. Indent and Format

   # Always use proper formatting
   json.dump(data, f, indent=4)
   

Validation Checklist

• All question keys present (q1, q2, q3)
• All answers are lists
• All token counts are integers
• No null values (use empty list [] instead)
• Output file is valid JSON
• Can be loaded with json.load()

Testing Output

import json

# Verify output file
with open('/root/answer.json', 'r') as f:
    data = json.load(f)

# Check structure
for q_key, q_data in data.items():
    assert "answer" in q_data
    assert "tokens" in q_data
    assert isinstance(q_data["answer"], list)
    assert isinstance(q_data["tokens"], int)
    print(f"{q_key}: {len(q_data['answer'])} items, {q_data['tokens']} tokens")