feat(azure,vertex): add 26 missing models for full provider coverage

[adaline-ankit]

Summary

• Azure: Add 25 missing OpenAI models to achieve near-full parity with the OpenAI provider
  • GPT-4.1 series: gpt-4.1, gpt-4.1-mini, gpt-4.1-nano
  • O-series: o1, o1-2024-12-17, o3, o3-2025-04-16, o3-mini, o3-mini-2025-01-31, o4-mini, o4-mini-2025-04-16
  • GPT-5 family: gpt-5, gpt-5.1, gpt-5.2, gpt-5.2-codex, gpt-5.2-chat-latest, gpt-5.3-codex, gpt-5-chat-latest, gpt-5-mini, gpt-5-nano, chatgpt-5.2
  • Remaining GPT-4 variants: gpt-4-turbo, gpt-4-turbo-preview, gpt-4-0125-preview, chatgpt-4o-latest
• Vertex: Add gemini-3.1-pro-preview-customtools with pricing, achieving parity with the Google provider

Note: gpt-5.2-pro is excluded from Azure — it uses the OpenAI Responses API (/v1/responses) which requires a new Azure base class (current Azure base only supports /chat/completions).

Test plan

• pnpm run build — 14/14 packages pass
• pnpm run test — 17/17 suites pass (Azure: 100 tests, Vertex: 83 tests)
• All new models follow existing Azure/Vertex patterns exactly
• Changesets created for both @adaline/azure (minor) and @adaline/vertex (minor)
• Verify Azure deployment-level auth works with o-series reasoning_effort param (requires live Azure endpoint)

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

This PR achieves near-full provider coverage by adding 25 missing OpenAI models to the Azure provider and one gemini-3.1-pro-preview-customtools variant to Vertex. All new files follow the established patterns precisely — Azure models delegate to BaseChatModelOpenAI with the corresponding schema imported from @adaline/openai, and the Vertex model mirrors the existing gemini-3.1-pro-preview structure with its own pricing entry and provider registration.

Key observations:

• GPT-4/GPT-5/ChatGPT variants (Azure): 17 new model files are structurally correct and will work with the current 2024-06-01 API version.
• O-series models (Azure): 8 new files (o1, o1-2024-12-17, o3, o3-2025-04-16, o3-mini, o3-mini-2025-01-31, o4-mini, o4-mini-2025-04-16) are structurally consistent but will likely fail at runtime — Azure's o-series endpoints require a minimum API version of 2024-12-01-preview, while BaseChatModelOpenAI hardcodes 2024-06-01. This gap is flagged by the unchecked test item in the PR description and should be resolved before the o-series additions can be considered production-ready.
• Vertex model: Correctly structured and fully registered in the provider factory; a minor specs-reference comment is missing compared to the sister gemini-3.1-pro-preview file.
• GitHub Actions workflow: The new claude.yml has broad write permissions (contents, PRs, issues) with no actor restriction, allowing any commenter to trigger Claude Code Action — worth tightening before merge.

Confidence Score: 3/5

• Safe to merge the GPT-4/GPT-5/Vertex changes; o-series Azure models and the GitHub Actions workflow each carry a functional/security concern that should be addressed first.
• The bulk of the PR (non-o-series Azure models + Vertex model) is clean and ready. However, the 8 new o-series Azure models will produce runtime failures against Azure endpoints due to the hardcoded 2024-06-01 API version — a real functional blocker for those specific additions. Additionally, the new GitHub Actions workflow grants write permissions without restricting who can trigger it, which is a security concern. These two issues together prevent a 4/5.
• core/providers/azure/src/models/chat-models/openai/base-chat-model.openai.azure.ts (API version) and .github/workflows/claude.yml (permission scope) need the most attention before merge.

Important Files Changed

Filename	Overview
core/providers/azure/src/models/chat-models/openai/base-chat-model.openai.azure.ts	Pre-existing hardcoded API version 2024-06-01 becomes a functional blocker for the newly added o-series models which require 2024-12-01-preview or newer on Azure.
core/providers/azure/src/models/chat-models/openai/index.ts	Barrel file updated to export all 25 new Azure model files in alphabetical order; consistent with existing export pattern.
core/providers/azure/src/models/chat-models/openai/o4-mini.openai.azure.ts	Representative of all 8 new o-series Azure model files; correctly delegates to BaseChatModelOpenAI with the corresponding OpenAI schema, but inherits the API version issue noted in the base class.
core/providers/vertex/src/models/chat-models/gemini-3.1-pro-preview-customtools.vertex.ts	New Vertex model follows the established pattern exactly (same config as the regular gemini-3.1-pro-preview); missing specs reference comment is a minor style gap.
core/providers/vertex/src/models/pricing.json	Pricing for gemini-3.1-pro-preview-customtools added with two token ranges and correct field structure matching all other entries.
core/providers/vertex/src/provider/provider.vertex.ts	New gemini-3.1-pro-preview-customtools entry correctly registered in the provider factory alongside its literal, options, and schema exports.
.github/workflows/claude.yml	New GitHub Actions workflow for Claude Code Action; grants contents: write and PR/issue write permissions without restricting which actors can trigger it, allowing any commenter to initiate write-access operations.
.claude/settings.json	Adds prettier/eslint/gh-cli permissions and PostToolUse/PreToolUse hooks; protected-file guard and auto-format hooks look correct.
core/providers/azure/src/models/chat-models/openai/gpt-4-1.openai.azure.ts	Representative of the 17 new non-o-series Azure model files; follows the existing pattern exactly with correct imports from @adaline/openai.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    User["User calls provider.chatModel(options)"]

    subgraph Azure ["Azure Provider"]
        AzureProv["Azure.chatModel()\n(always uses __base__)"]
        BaseChatModel["BaseChatModelOpenAI\napi-version: 2024-06-01 (hardcoded)"]
        AzureProv --> BaseChatModel
    end

    subgraph NewAzureModels ["New Model Schemas (exported, not registered)"]
        GPT41["GPT-4.1 / GPT-4.1-mini / GPT-4.1-nano\n✅ compatible with 2024-06-01"]
        GPT5["GPT-5 family (10 models)\n✅ compatible with 2024-06-01"]
        GPT4Turbo["GPT-4-turbo / GPT-4-0125-preview\n✅ compatible with 2024-06-01"]
        OSeries["o1, o3, o4-mini + dated variants (8 models)\n⚠️ requires ≥ 2024-12-01-preview"]
    end

    User --> AzureProv
    BaseChatModel -.->|"schema reference only\n(not wired into provider)"| GPT41
    BaseChatModel -.->|"schema reference only"| GPT5
    BaseChatModel -.->|"schema reference only"| GPT4Turbo
    BaseChatModel -.->|"schema reference only\n❌ API version mismatch"| OSeries

    subgraph Vertex ["Vertex Provider"]
        VertexProv["Vertex.chatModel(options)"]
        GeminiFactory["Model factory lookup by literal"]
        NewGemini["gemini-3.1-pro-preview-customtools\n✅ registered + priced"]
        VertexProv --> GeminiFactory --> NewGemini
    end

    User --> VertexProv

Loading

Comments Outside Diff (2)

1. .github/workflows/claude.yml, line 141-144 (link)

   No actor restriction on Claude trigger

   The workflow grants contents: write, pull-requests: write, and issues: write permissions and fires for any comment containing `@claude` — including comments from external contributors or bots on public repositories. Without an if: github.actor == '...' guard or a check that the comment author is a collaborator/member of the org, any user who can post a comment can trigger Claude Code Action with write access.

   Consider adding an actor check, e.g.:

   Or check github.event.comment.author_association against 'OWNER', 'MEMBER', or 'COLLABORATOR' to restrict trigger to trusted contributors.

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: .github/workflows/claude.yml
   Line: 141-144
   
   Comment:
   **No actor restriction on Claude trigger**
   
   The workflow grants `contents: write`, `pull-requests: write`, and `issues: write` permissions and fires for any comment containing `` `@claude` `` — including comments from external contributors or bots on public repositories. Without an `if: github.actor == '...'` guard or a check that the comment author is a collaborator/member of the org, any user who can post a comment can trigger Claude Code Action with write access.
   
   Consider adding an actor check, e.g.:
   
   
   
   Or check `github.event.comment.author_association` against `'OWNER'`, `'MEMBER'`, or `'COLLABORATOR'` to restrict trigger to trusted contributors.
   
   How can I resolve this? If you propose a fix, please make it concise.

2. core/providers/azure/src/models/chat-models/openai/base-chat-model.openai.azure.ts, line 31 (link)

   Hardcoded API version incompatible with o-series models

   The BaseChatModelOpenAI constructor hardcodes api-version=2024-06-01. All 8 new o-series models added in this PR (o1, o1-2024-12-17, o3, o3-2025-04-16, o3-mini, o3-mini-2025-01-31, o4-mini, o4-mini-2025-04-16) inherit from this class. Azure's o-series models require 2024-12-01-preview or newer to support the updated request schema and the reasoning_effort parameter — requests using the 2024-06-01 version will likely fail with a 400 or 404 from the Azure endpoint.

   This is a pre-existing base-class limitation, but it becomes a real functional blocker specifically for these new o-series additions (the unchecked test item in the PR description is a direct symptom of this gap).

   The cleanest fix would be to allow the API version to be injected into the constructor as an optional parameter (defaulting to "2024-06-01" for backward compatibility), then have each o-series model class pass "2025-01-01-preview" when calling super. That way existing GPT-4/GPT-4o models stay unaffected while the reasoning models use the correct version.

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: core/providers/azure/src/models/chat-models/openai/base-chat-model.openai.azure.ts
   Line: 31
   
   Comment:
   **Hardcoded API version incompatible with o-series models**
   
   The `BaseChatModelOpenAI` constructor hardcodes `api-version=2024-06-01`. All 8 new o-series models added in this PR (`o1`, `o1-2024-12-17`, `o3`, `o3-2025-04-16`, `o3-mini`, `o3-mini-2025-01-31`, `o4-mini`, `o4-mini-2025-04-16`) inherit from this class. Azure's o-series models require **`2024-12-01-preview`** or newer to support the updated request schema and the `reasoning_effort` parameter — requests using the `2024-06-01` version will likely fail with a `400` or `404` from the Azure endpoint.
   
   This is a pre-existing base-class limitation, but it becomes a real functional blocker specifically for these new o-series additions (the unchecked test item in the PR description is a direct symptom of this gap).
   
   The cleanest fix would be to allow the API version to be injected into the constructor as an optional parameter (defaulting to `"2024-06-01"` for backward compatibility), then have each o-series model class pass `"2025-01-01-preview"` when calling `super`. That way existing GPT-4/GPT-4o models stay unaffected while the reasoning models use the correct version.
   
   How can I resolve this? If you propose a fix, please make it concise.

Prompt To Fix All With AI

This is a comment left during a code review.
Path: .github/workflows/claude.yml
Line: 141-144

Comment:
**No actor restriction on Claude trigger**

The workflow grants `contents: write`, `pull-requests: write`, and `issues: write` permissions and fires for any comment containing `` `@claude` `` — including comments from external contributors or bots on public repositories. Without an `if: github.actor == '...'` guard or a check that the comment author is a collaborator/member of the org, any user who can post a comment can trigger Claude Code Action with write access.

Consider adding an actor check, e.g.:

```suggestion
    if: |
      github.actor == 'adaline-ankit' || contains(fromJson('["owner", "member", "collaborator"]'), github.event.comment.author_association)
      && (
        (github.event_name == 'issue_comment' && contains(github.event.comment.body, '@claude')) ||
        (github.event_name == 'pull_request_review_comment' && contains(github.event.comment.body, '@claude')) ||
        (github.event_name == 'issues' && contains(github.event.issue.labels.*.name, 'claude'))
      )
```

Or check `github.event.comment.author_association` against `'OWNER'`, `'MEMBER'`, or `'COLLABORATOR'` to restrict trigger to trusted contributors.

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: core/providers/vertex/src/models/chat-models/gemini-3.1-pro-preview-customtools.vertex.ts
Line: 14

Comment:
**Missing specs reference comment**

The sister file `gemini-3.1-pro-preview.vertex.ts` includes a `// Specs reference: https://cloud.google.com/vertex-ai/generative-ai/docs/models/model-versions (retrieved 2026-02-25)` comment above the literal constant for traceability. The new `customtools` variant omits this. Adding it (with the source and retrieval date for the pricing/capacity numbers) would keep the file consistent with the established pattern in this directory.

```suggestion
// Specs reference: https://cloud.google.com/vertex-ai/generative-ai/docs/models/model-versions (retrieved <date>)
const Gemini3_1ProPreviewCustomtoolsLiteral = "gemini-3.1-pro-preview-customtools" as const;
```

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: core/providers/azure/src/models/chat-models/openai/base-chat-model.openai.azure.ts
Line: 31

Comment:
**Hardcoded API version incompatible with o-series models**

The `BaseChatModelOpenAI` constructor hardcodes `api-version=2024-06-01`. All 8 new o-series models added in this PR (`o1`, `o1-2024-12-17`, `o3`, `o3-2025-04-16`, `o3-mini`, `o3-mini-2025-01-31`, `o4-mini`, `o4-mini-2025-04-16`) inherit from this class. Azure's o-series models require **`2024-12-01-preview`** or newer to support the updated request schema and the `reasoning_effort` parameter — requests using the `2024-06-01` version will likely fail with a `400` or `404` from the Azure endpoint.

This is a pre-existing base-class limitation, but it becomes a real functional blocker specifically for these new o-series additions (the unchecked test item in the PR description is a direct symptom of this gap).

The cleanest fix would be to allow the API version to be injected into the constructor as an optional parameter (defaulting to `"2024-06-01"` for backward compatibility), then have each o-series model class pass `"2025-01-01-preview"` when calling `super`. That way existing GPT-4/GPT-4o models stay unaffected while the reasoning models use the correct version.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "feat(azure,vertex): add 26 missing model..." | Re-trigger Greptile}

[greptile-apps]

Missing specs reference comment

The sister file gemini-3.1-pro-preview.vertex.ts includes a // Specs reference: https://cloud.google.com/vertex-ai/generative-ai/docs/models/model-versions (retrieved 2026-02-25) comment above the literal constant for traceability. The new customtools variant omits this. Adding it (with the source and retrieval date for the pricing/capacity numbers) would keep the file consistent with the established pattern in this directory.

Suggested change

	// Specs reference: https://cloud.google.com/vertex-ai/generative-ai/docs/models/model-versions (retrieved <date>)
	const Gemini3_1ProPreviewCustomtoolsLiteral = "gemini-3.1-pro-preview-customtools" as const;

Prompt To Fix With AI

This is a comment left during a code review.
Path: core/providers/vertex/src/models/chat-models/gemini-3.1-pro-preview-customtools.vertex.ts
Line: 14

Comment:
**Missing specs reference comment**

The sister file `gemini-3.1-pro-preview.vertex.ts` includes a `// Specs reference: https://cloud.google.com/vertex-ai/generative-ai/docs/models/model-versions (retrieved 2026-02-25)` comment above the literal constant for traceability. The new `customtools` variant omits this. Adding it (with the source and retrieval date for the pricing/capacity numbers) would keep the file consistent with the established pattern in this directory.

```suggestion
// Specs reference: https://cloud.google.com/vertex-ai/generative-ai/docs/models/model-versions (retrieved <date>)
const Gemini3_1ProPreviewCustomtoolsLiteral = "gemini-3.1-pro-preview-customtools" as const;
```

How can I resolve this? If you propose a fix, please make it concise.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}