Add ChatGPT Agents SDK and Amazon Bedrock AgentCore cookbook - #3057

Summary

Adds an end-to-end Cookbook example for building a private ChatGPT flight-assistant plugin with the OpenAI Agents SDK and Amazon Bedrock AgentCore.

The contribution includes:

• a runnable notebook and self-contained setup guide;
• a validated, read-only MCP adapter and ChatGPT widget;
• local and optional deployed AgentCore Runtime execution paths;
• Secure MCP Tunnel guidance;
• AWS-default and optional dual tracing;
• credential-free Promptfoo evaluations and an optional AWS-native evaluation path;
• least-privilege IAM and publication-boundary documentation.

The content is isolated under examples/partners/AWS/chatgpt_agents_sdk_aws_agentcore_cookbook/. A matching registry.yaml entry publishes the notebook on the Cookbook site.

Motivation

Enterprise developers often need to connect ChatGPT to privately hosted tools without exposing a public MCP endpoint or assuming broad AWS permissions. This example provides a reproducible, least-privilege workflow with explicit credential, tracing, evaluation, and operational boundaries.

Validation

• 131 Python tests passed.
• Ruff check and format checks passed.
• Pyright passed with 0 errors.
• 69 MCP adapter tests passed.
• MCP adapter build and TypeScript typecheck passed.
• 58 Promptfoo runner/assertion tests passed.
• Promptfoo offline and actual-agent configurations validated.
• 15 credential-free Promptfoo evaluation cases passed.
• The source notebook executed top to bottom with live AWS and agent-evaluation gates disabled.
• The OpenAI Cookbook notebook validator passed.
• Registry metadata parses and resolves to the new notebook.
• Local Markdown and notebook links resolve.
• The staged publication tree passed credential and private-path scans.

Live AWS, OpenAI trace export, and hosted Secure MCP Tunnel calls were not rerun as part of this public-repository copy; those paths remain explicitly opt-in and credential-gated.

Risk and rollback

The change adds a self-contained example plus one registry entry and does not modify existing examples. It can be rolled back by reverting the contribution commit.

For new content

• I have added a new entry in registry.yaml (and, optionally, in authors.yaml) so that my content renders on the cookbook website.
• I have conducted a self-review of my content based on the contribution guidelines:
  • Relevance: This content is related to building with OpenAI technologies and is useful to others.
  • Uniqueness: I reviewed the existing AWS partner examples and verified that this contribution adds a distinct ChatGPT, Secure MCP Tunnel, tracing, and evaluation workflow.
  • Spelling and Grammar: I reviewed the changed Markdown and notebook Markdown cells with the repository's docs-editor checklist.
  • Clarity: I completed a final read-through and verified that the submission is organized as an end-to-end workflow.
  • Correctness: Credential-free tests and notebook execution pass; credentialed paths are clearly labeled and gated.
  • Completeness: Required setup, environment variables, security boundaries, validation, and references are documented.