feat: add openai-chatgpt provider to routing matrix#18
Open
robotdad wants to merge 2 commits into
Open
Conversation
Add the ChatGPT subscription backend (via Codex CLI OAuth) to the routing matrix bundle: - New file: routing/openai-chatgpt.yaml — single-provider matrix covering all 13 roles Uses gpt-5.5 as primary with gpt-5.4 fallbacks, codex-tuned models for coding role. No pro tier—reasoning roles use gpt-5.5 + reasoning_effort:high instead. - Updated: routing/balanced.yaml — Added openai-chatgpt candidates after each openai entry (13 entries total). Maintains ordering convention (anthropic > openai > openai-chatgpt > gemini > copilot > ollama). - Updated: routing/quality.yaml — Same pattern, 12 entries (no openai entry in image-gen to follow). - Updated: routing/economy.yaml — Same pattern, 12 entries. Maps openai nano tier to chatgpt mini. Budget roles use gpt-?.?-mini*, quality-sensitive roles use gpt-5.5. - Updated: docs/MATRIX_CURATOR_GUIDE.md — Added openai-chatgpt to the provider naming table. Provider module lives at microsoft/amplifier-module-provider-openai-chatgpt. Uses Codex CLI's OAuth device code flow (no API key needed—works with ChatGPT Plus/Pro/Team/Enterprise). Generated with Amplifier Co-Authored-By: Amplifier <240397093+microsoft-amplifier@users.noreply.github.com>
Add a Digital Twin Universe profile for end-to-end validation of the routing matrix with the openai-chatgpt provider. This profile provisions an isolated container with Amplifier CLI, the chatgpt provider module, and the routing matrix bundle, then verifies all 13 roles route correctly and live inference succeeds through the ChatGPT backend. All 20 acceptance tests pass (14 MUST, 4 SHOULD, 2 NICE): - All 13 roles route to openai-chatgpt with correct models - Live inference through ChatGPT backend confirmed (gpt-5.5, 5.4s response) - All 8 routing YAML files parse cleanly with required fields - Candidate ordering rule (anthropic > openai > openai-chatgpt > gemini > copilot) - No duplicate chatgpt entries in any role - image-gen correctly excluded from multi-provider matrices - Every role in single-provider matrix has gpt-5.4 fallback Generated with Amplifier Co-Authored-By: Amplifier <240397093+microsoft-amplifier@users.noreply.github.com>
robotdad
added a commit
to microsoft/amplifier
that referenced
this pull request
Apr 30, 2026
Adds openai-chatgpt provider entry to the official Amplifier module catalog in MODULES.md. This provider connects to ChatGPT subscriptions (Plus, Pro, Team, Enterprise) via Codex CLI's OAuth device code flow. No API key required — authenticates using the same mechanism as the Codex CLI itself, supporting the full set of ChatGPT models via the subscription endpoint. Related: microsoft/amplifier-bundle-routing-matrix#18 Co-Authored-By: Amplifier <240397093+microsoft-amplifier@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The PR adds the
openai-chatgptprovider (ChatGPT subscription backend via Codex CLI OAuth device code flow) to the routing matrix bundle. The provider module lives atmicrosoft/amplifier-module-provider-openai-chatgpt.Changes
routing/openai-chatgpt.yaml— Complete single-provider matrix covering all 13 roles. Uses gpt-5.5 as primary with gpt-5.4 fallbacks. Codex-tuned models (gpt-?.?-codex*) preferred for coding role. No pro tier — reasoning roles use gpt-5.5 + reasoning_effort:high.routing/balanced.yaml— Added openai-chatgpt candidates after each openai entry (13 entries). Ordering: anthropic > openai > openai-chatgpt > gemini > copilot > ollama.routing/quality.yaml— Same pattern, 12 entries (image-gen has no openai entry to follow).routing/economy.yaml— Same pattern, 12 entries. Maps openai's nano tier to chatgpt's mini (no nano available).docs/MATRIX_CURATOR_GUIDE.md— Addedopenai-chatgpt→provider-openai-chatgptto provider naming table..amplifier/digital-twin-universe/profiles/routing-matrix-chatgpt-reality-check.yaml— DTU profile for end-to-end validation.About the Provider
openai-chatgptgpt-?.?-pro*tier — reasoning handled via reasoning_effort dial on gpt-5.5DTU Validation (20/20 tests passed)
A Digital Twin Universe environment was provisioned and all acceptance tests passed:
Key verifications:
openai-chatgptwith correct modelsKnown Issue (not from this PR)
hooks-routing resolves matrix files under its own bundle cache hash (upstream main) rather than the active app bundle's hash. Inference still works via provider defaults, but
amplifier routing showmay report "routing disabled" when the matrix bundle is registered from a fork/branch. This is an existing hooks-routing module issue.