feat(openai)!: rewrite plugin for 1.0

[WH-2099]

Summary

Fixes #3418.

• Rewrite the OpenAI plugin runtime around the official OpenAI Python SDK with small, protocol-focused modules.
• Make the Responses API the default for supported language models while retaining Chat Completions for audio models and compatible endpoints.
• Preserve stateless reasoning by storing and replaying complete output items, encrypted reasoning content, and assistant phase metadata in order.
• Request encrypted reasoning only from GPT-5 and o-series reasoning models, because non-reasoning Responses models reject that include field.
• Handle reasoning summaries, refusals, parallel function calls, interleaved streaming events, stop sequences, incomplete responses, errors, usage, and stream closure explicitly.
• Simplify embeddings, moderation, transcription, and speech generation while removing direct NumPy and pydub dependencies.
• Audit every model configuration against the official model catalog, model pages, and deprecation ledger.
• Remove the deprecated gpt-4o, gpt-audio-mini, and gpt-4o-mini-tts aliases while retaining their still-current dated snapshots.
• Prevent all six deprecated GPT-5/o3 dated snapshots announced on 2026-06-11 from returning to the catalog.
• Align limits, reasoning controls, service tiers, prices, voices, file formats, and model capabilities with current OpenAI documentation.
• Replace the scattered legacy tests with parameterized pytest suites using pytest-mock.
• Add a live API matrix derived from every presented LLM and every non-LLM model configuration.
• Load live-test credentials from an ignored .env, redact them in pytest output, run automatically for a nonempty key, and dynamically skip otherwise.
• Remove the unrelated remote-debug .env.example and document the billable workflow in the README.
• Rewrite the README and bump the plugin to 1.0.0.

Breaking changes

• The plugin now defaults supported LLMs to the Responses API.
• Deprecated and unavailable model configurations are removed.
• The deprecated gpt-4o, gpt-audio-mini, and gpt-4o-mini-tts aliases are no longer presented.
• Empty embedding and TTS inputs now fail with explicit bad-request errors.
• The legacy Completions path and obsolete duplicate integration-test harness are removed.

Change Type

• Documentation / non-plugin change
• Non-LLM plugin only
• LLM and multimodal model plugin

LLM Plugin Checklist

Areas affected by this change

• System, developer, user, assistant, and tool message flow
• Multi-round and parallel tool interaction
• Image, file, and audio inputs
• Structured output
• Reasoning summaries and stateless reasoning replay
• Token usage and pricing metadata
• Streaming order, stop handling, terminal states, errors, and cancellation
• Model catalog and parameter rules

Version

• manifest.yaml and pyproject.toml are 1.0.0.
• openai>=2.45.0 and dify_plugin>=0.9.0 are declared and locked.
• pytest-mock>=3.15.1 is included in the development dependency group.

Testing

• Python 3.12 full suite with the configured .env: 210 passed, 5 skipped.
• All 63 live cases run automatically for a nonempty OPENAI_API_KEY and skip when the key is missing, empty, or whitespace-only.
• The current live suite collects 38 presented LLMs, 12 non-LLM configurations, and 13 representative protocol and capability boundaries.
• The complete current live matrix was covered serially in cost-controlled batches: 58 passed and 5 skipped.
• Four presented aliases were skipped because this API organization requires verification for gpt-5, gpt-5-mini, o3, and o3-pro.
• The reasoning-summary case was skipped because OpenAI accepted the request but withheld the summary from the unverified organization.
• uv run pytest --strict-markers tests/live -q
• uv run ruff check .
• uv run ruff format --check .
• uv lock --check
• git diff --check

Official references

• https://developers.openai.com/api/docs/models
• https://developers.openai.com/api/docs/models/gpt-4o
• https://developers.openai.com/api/docs/models/gpt-audio-mini
• https://developers.openai.com/api/docs/models/gpt-4o-mini-tts
• https://developers.openai.com/api/docs/guides/latest-model
• https://developers.openai.com/api/docs/guides/migrate-to-responses
• https://developers.openai.com/api/docs/guides/reasoning
• https://developers.openai.com/api/docs/guides/function-calling
• https://developers.openai.com/api/docs/guides/file-inputs
• https://developers.openai.com/api/docs/guides/speech-to-text
• https://developers.openai.com/api/docs/guides/text-to-speech
• https://developers.openai.com/api/docs/deprecations
• https://developers.openai.com/api/docs/pricing

[gemini-code-assist]

Note

Gemini is unable to generate a review for this pull request due to the file types involved not being currently supported.

[Copilot]

Pull request overview

Adds OpenAI GPT-5.6 model-family entries to the OpenAI plugin’s predefined LLM model catalog, so the models appear in Dify’s model list with appropriate limits/capabilities and an updated plugin version.

Changes:

• Added YAML metadata for gpt-5.6 (alias) plus gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna.
• Updated the LLM model ordering list to include the new GPT-5.6 entries.
• Bumped the OpenAI plugin manifest version to 0.4.4.

Reviewed changes

Copilot reviewed 6 out of 7 changed files in this pull request and generated 10 comments.

Show a summary per file

File	Description
models/openai/models/llm/gpt-5.6.yaml	Adds the gpt-5.6 alias model definition (features/params/pricing).
models/openai/models/llm/gpt-5.6-sol.yaml	Adds GPT-5.6 Sol model definition.
models/openai/models/llm/gpt-5.6-terra.yaml	Adds GPT-5.6 Terra model definition.
models/openai/models/llm/gpt-5.6-luna.yaml	Adds GPT-5.6 Luna model definition.
models/openai/models/llm/_position.yaml	Inserts the new GPT-5.6 entries into the model ordering list.
models/openai/manifest.yaml	Bumps plugin version from 0.4.3 to 0.4.4.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.