[🐛 Fix]: Upgrade deprecated OpenAI API client and replace davinci model with gpt-3.5-turbo

[lucumango]

Description 📝

The Synthesize.ipynb research notebook uses the deprecated openai==0.28 Python client and the davinci completion model, which are no longer supported by OpenAI and throw InvalidRequestError / deprecation warnings when researchers attempt to run the data synthesis pipeline.

Problem 🚨

• openai==0.28 is deprecated and incompatible with the current OpenAI API.
• openai.Completion.create with the davinci engine is no longer available.
• The hardcoded openai.api_key assignment is a security anti-pattern.
• No error handling around API calls causes the entire batch loop to crash on quota/rate-limit errors.

Proposed Solution 💡

1. Upgrade the pip dependency to openai>=1.0.0.
2. Initialize the client securely using os.environ.get('OPENAI_API_KEY') instead of hardcoding the key.
3. Replace the deprecated openai.Completion.create / davinci call with the modern client.chat.completions.create syntax using gpt-3.5-turbo.
4. Wrap API calls in a try/except block to gracefully handle rate-limit or quota errors without crashing the batch loop.

References 🔗

• OpenAI API Deprecations

Acceptance Criteria ✅

• Synthesize.ipynb runs without deprecation warnings or InvalidRequestError.
• API key is read from environment variables, not hardcoded.
• Text generation uses gpt-3.5-turbo via client.chat.completions.create.
• Batch loop handles rate-limit errors gracefully.