examples(elasticsearch): migrate notebook to openai>=1.0 client API

[kiwigitops]

Summary

examples/vector_databases/elasticsearch/elasticsearch-retrieval-augmented-generation.ipynb still uses the pre-v1 OpenAI Python SDK (openai.api_key = ..., openai.Embedding.create(...), openai.ChatCompletion.create(...), dict-style response["data"][0]["embedding"]). The pre-v1 API was removed in openai>=1.0.0 (Nov 2023), so the notebook now fails with AttributeError: module 'openai' has no attribute 'Embedding' on the current SDK.

This PR migrates the affected cells. The Elasticsearch client variable is left alone — the new OpenAI client is named openai_client to avoid collision.

Changes

-import openai
+from openai import OpenAI

-openai.api_key = OPENAI_API_KEY
+openai_client = OpenAI(api_key=OPENAI_API_KEY)
...
-question_embedding = openai.Embedding.create(input=question, model=EMBEDDING_MODEL)
+question_embedding = openai_client.embeddings.create(input=question, model=EMBEDDING_MODEL)

-"query_vector": question_embedding["data"][0]["embedding"],
+"query_vector": question_embedding.data[0].embedding,

-summary = openai.ChatCompletion.create(
+summary = openai_client.chat.completions.create(
   model="gpt-3.5-turbo", messages=[...]
 )

The downstream usage summary.choices/choice.message.content already works on the v1+ response shape, so only the create call needed swapping.

Test plan

• Notebook JSON validates
• No remaining openai.ChatCompletion, openai.Embedding, openai.api_key, or dict-style ["data"][0]["embedding"] access in code cells
• Same model, same parameters — purely SDK-surface migration