Practical Tips for ChatGPT API Integration

Recently, I helped several clients integrate ChatGPT and summarized some practical experiences to share with you.

1. Control Token Usage

API calls are billed by tokens, so controlling usage is important:

# Limit response length
response = openai.ChatCompletion.create(
    model="gpt-3.5-turbo",
    messages=messages,
    max_tokens=500  # Limit max tokens
)

2. Set Reasonable Timeout

API sometimes responds slowly, it's recommended to set a timeout:

import openai
openai.api_request_timeout = 30  # 30 seconds timeout

3. Handle Errors Properly

Network issues, rate limiting, etc. may cause call failures:

try:
    response = openai.ChatCompletion.create(...)
except openai.error.RateLimitError:
    # Rate limit triggered, retry later
    time.sleep(60)
except openai.error.APIError:
    # API error, log it
    logging.error("API call failed")

4. Use Streaming for Better UX

Long waiting times lead to poor experience. Use streaming:

response = openai.ChatCompletion.create(
    model="gpt-3.5-turbo",
    messages=messages,
    stream=True  # Enable streaming
)

for chunk in response:
    print(chunk.choices[0].delta.content, end="")

5. Content Filtering

Make sure to do content moderation before going live to avoid inappropriate content.

These are some basic experiences. For more details, feel free to contact us.