feat(learn): modernize gen-qa-openai notebook to OpenAI SDK v1.x

[jhamon]

Problem

The learn/generation/openai/gen-qa-openai.ipynb notebook uses deprecated OpenAI SDK v0.27 with the legacy Completion API and the retired text-davinci-003 model. This prevents users from running the notebook and demonstrates outdated patterns.

Solution

Modernized the notebook to use OpenAI SDK v1.66.3 with the current Chat Completions API and gpt-5-mini model. This brings the notebook up to date with current best practices and ensures it will continue working with OpenAI's supported model lineup.

Key Changes

1. Dependencies: Updated openai==0.27.7 → openai==1.66.3 and pinecone-client[grpc]==2.2.1 → pinecone==8.0.0

2. API Initialization: Migrated from global configuration to client-based pattern:

   # Before
   import openai
   openai.api_key = os.getenv("OPENAI_API_KEY")
   
   # After
   from openai import OpenAI
   client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

3. Chat Completions API: Replaced Completion API with Chat API using gpt-5-mini:

   # Before
   res = openai.Completion.create(
       engine='text-davinci-003',
       prompt=query,
       temperature=0,
       max_tokens=400
   )
   res['choices'][0]['text'].strip()
   
   # After
   res = client.chat.completions.create(
       model='gpt-5-mini',
       messages=[{"role": "user", "content": query}],
       temperature=0,
       max_tokens=400
   )
   res.choices[0].message.content.strip()

4. Embeddings API: Updated to new client method and response access patterns:

   # Before
   res = openai.Embedding.create(input=[text], engine='text-embedding-ada-002')
   embeds = [record['embedding'] for record in res['data']]
   
   # After
   res = client.embeddings.create(input=[text], model='text-embedding-ada-002')
   embeds = [record.embedding for record in res.data]

Model Selection

Selected gpt-5-mini because:

• Part of OpenAI's current GPT-5 flagship series (as of Feb 2026)
• Cost-efficient and well-suited for focused RAG tasks
• Not being deprecated, ensuring notebook longevity
• More capable than retiring GPT-4.1/GPT-4o models

Breaking Changes

None for end users - the notebook functionality remains the same. API keys and usage patterns are maintained.

Testing

Verified through code pattern analysis:

• ✅ All legacy API patterns removed (openai.Completion, openai.Embedding)
• ✅ Modern client-based patterns present throughout
• ✅ Correct model (gpt-5-mini) and response access patterns
• ✅ All 5 embedding API calls updated
• ✅ Both complete() and retrieve() functions modernized

Full notebook execution requires valid OpenAI and Pinecone API keys.

Related

Related to other SDK v8 modernization efforts across the examples repository.

Made with Cursor

---

Note

Low Risk
Low risk since changes are limited to tutorial notebooks, but execution could break if the pinned OpenAI/Pinecone SDK versions or chosen models differ from a user’s environment.

Overview
Modernizes the OpenAI RAG learning notebooks to current SDKs. Updates pinned deps to OpenAI SDK v1 and Pinecone v8, and switches OpenAI usage from global openai.* calls to the client-based OpenAI(...) pattern.

Moves generation to chat-based models and updates response handling. gen-qa-openai.ipynb replaces text-davinci-003 completions with client.chat.completions.create(model="gpt-5-mini"), while openai-ml-qa/01-making-queries.ipynb uses chat completions with gpt-3.5-turbo; both notebooks update embeddings calls to client.embeddings.create(..., model=...) and migrate all response indexing from dict-style (res['data'], res['choices']) to attribute access (res.data, res.choices).

^{Written by Cursor Bugbot for commit a6bbb78. This will update automatically on new commits. Configure here.}

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[jhamon]

Closing in favor of #546 which contains only the gen-qa-openai.ipynb changes without unrelated commits from the previous branch.