Skip to content

07. External APIs & data manipulation

🎯 Objective

Apply OpenCode to real integration tasks: consume REST APIs, handle errors, and process data — with the agent writing, running, and explaining the code.

📋 Consuming an API with code

Ask OpenCode to write and test the call directly in your project. Study example in exemplos/api_exemplo.py:

import requests

def fetch_data(url, token=None):
    headers = {"Authorization": f"Bearer {token}"} if token else {}
    try:
        response = requests.get(url, headers=headers, timeout=10)
        response.raise_for_status()
        return response.json()
    except requests.exceptions.RequestException as e:
        print(f"Request error: {e}")
        return None

Best practices: - Always set a timeout + try/except for network errors - Validate status_code before parsing - Keep API keys in environment variables — never in code - Limit the output: use opencode run --format json for pipelines

📋 Data manipulation

See exemplos/dados_exemplo.py — builds a formatted JSON report from a list of grades:

import json
from datetime import datetime

students = json.loads(raw_data)
report = {
    "total_students": len(students),
    "overall_average": sum(s["grade"] for s in students) / len(students),
    "highest": max(students, key=lambda x: x["grade"]),
}
print(json.dumps(report, indent=2, ensure_ascii=False))

With OpenCode in the flow

# run and explain the script
opencode run --file exemplos/dados_exemplo.py "Run and explain the generated report"

# structured output for data pipelines
opencode run --format json --file data.json "Summarize this dataset in 3 metrics"

🧠 Prompt tips for data

  • Pass the file with --file or @file instead of pasting the content
  • Ask for the output format explicitly (JSON, table, CSV)
  • For large inputs, ask for paging/incremental analysis (read pages by lines)

❓ Hands-on challenge

  1. Ask OpenCode: "create a function that fetches the first 10 posts from https://jsonplaceholder.typicode.com and prints each title"
  2. Add error handling and a timeout, then test with an invalid URL
  3. Ask to extend dados_exemplo.py with the median and the best student by average
  4. Produce the output with --format json and parse it with jq

📚 Official source

Next step: Module 03 — Advanced 🇧🇷 · Back to module