feat(azure): added azure anthropic, added backwards compat support for chat completions API, added opus 4.6

[waleedlatif1]

Summary

• added azure anthropic, added backwards compat support for chat completions API, added opus 4.6

Type of Change

• New feature

Testing

Tested manually

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		Feb 5, 2026 6:49pm

[greptile-apps]

Greptile Overview

Greptile Summary

This PR adds three major features:

1. Azure Anthropic Provider

• Adds new azure-anthropic provider that enables Claude models via Azure AI Foundry
• Extracts shared Anthropic logic into anthropic/core.ts for code reuse between standard and Azure providers
• Azure provider properly configures baseURL ({endpoint}/anthropic) and headers (api-key, anthropic-version)
• Includes 5 Azure Anthropic models: Claude Opus 4.6, Opus 4.5, Opus 4.1, Sonnet 4.5, and Haiku 4.5

2. Claude Opus 4.6 with Adaptive Thinking

• Adds Claude Opus 4.6 model definition with support for "max" thinking level (4 levels: low, medium, high, max)
• Implements adaptive thinking for Opus 4.6 using thinking: {type: 'adaptive'} + output_config: {effort} pattern
• Maintains backwards compatibility with budget-based extended thinking for older models (Opus 4.5, 4.1, 4.0, Sonnet, Haiku)
• Updates agent block UI to support max thinking level

3. Azure OpenAI Chat Completions Backwards Compatibility

• Adds URL parsing utilities to detect and handle full chat completions endpoint URLs
• Extracts deployment name, api-version, and base URL from full endpoint strings
• Enables users to pass complete Azure URLs instead of just base endpoints
• Falls back gracefully to existing Responses API behavior when no chat completions pattern detected

Confidence Score: 5/5

• This PR is safe to merge with no critical issues identified
• Well-structured implementation with proper separation of concerns. The shared core logic pattern is clean and maintainable. All type definitions are properly updated. The adaptive thinking implementation correctly distinguishes between Opus 4.6 and older models. Backwards compatibility additions are non-breaking and gracefully fallback.
• No files require special attention

Important Files Changed

Filename	Overview
apps/sim/providers/anthropic/core.ts	New shared core implementation for Anthropic providers with thinking capability support (Opus 4.6 adaptive thinking + budget-based extended thinking)
apps/sim/providers/azure-anthropic/index.ts	New Azure Anthropic provider using shared core implementation with Azure-specific client configuration
apps/sim/providers/models.ts	Added Claude Opus 4.6 model definition, Azure Anthropic provider models, and max thinking level support
apps/sim/providers/azure-openai/index.ts	Added backwards compatibility for chat completions API endpoint format
apps/sim/providers/azure-openai/utils.ts	Added utility functions for parsing chat completions endpoint URLs (extracting base URL, deployment, api-version)

Sequence Diagram

sequenceDiagram
    participant Client
    participant ProviderRegistry
    participant AzureAnthropicProvider
    participant AnthropicProvider
    participant CoreImplementation
    participant AnthropicSDK
    participant AzureEndpoint

    Note over Client,AzureEndpoint: Azure Anthropic Provider Flow (New)
    
    Client->>ProviderRegistry: executeRequest(model: azure-anthropic/claude-opus-4-6)
    ProviderRegistry->>AzureAnthropicProvider: executeRequest()
    
    AzureAnthropicProvider->>AzureAnthropicProvider: Validate azureEndpoint & apiKey
    AzureAnthropicProvider->>AzureAnthropicProvider: Strip azure-anthropic/ prefix from model
    AzureAnthropicProvider->>AzureAnthropicProvider: Construct baseURL: {endpoint}/anthropic
    
    AzureAnthropicProvider->>CoreImplementation: executeAnthropicProviderRequest(request, config)
    
    Note over CoreImplementation: Shared Core Logic
    
    CoreImplementation->>CoreImplementation: Check thinkingLevel capability
    alt Opus 4.6 (adaptive thinking)
        CoreImplementation->>CoreImplementation: Set thinking: {type: 'adaptive'}, output_config: {effort}
    else Other models (extended thinking)
        CoreImplementation->>CoreImplementation: Set thinking: {type: 'enabled', budget_tokens}
    end
    
    CoreImplementation->>AnthropicSDK: Create client with Azure baseURL + headers
    CoreImplementation->>AzureEndpoint: POST /anthropic/v1/messages
    AzureEndpoint-->>CoreImplementation: Response with tool calls
    
    alt Tools present
        CoreImplementation->>CoreImplementation: Execute tools
        CoreImplementation->>AzureEndpoint: Continue conversation with tool results
    end
    
    AzureEndpoint-->>CoreImplementation: Final response
    CoreImplementation-->>AzureAnthropicProvider: ProviderResponse
    AzureAnthropicProvider-->>ProviderRegistry: ProviderResponse
    ProviderRegistry-->>Client: Result

    Note over Client,AnthropicSDK: Standard Anthropic Provider Flow (Refactored)
    
    Client->>ProviderRegistry: executeRequest(model: claude-opus-4-6)
    ProviderRegistry->>AnthropicProvider: executeRequest()
    AnthropicProvider->>CoreImplementation: executeAnthropicProviderRequest(request, config)
    CoreImplementation->>AnthropicSDK: Create client with standard config
    CoreImplementation->>AnthropicSDK: POST /v1/messages
    AnthropicSDK-->>CoreImplementation: Response
    CoreImplementation-->>Client: Result

    Note over Client,AzureEndpoint: Azure OpenAI Backwards Compatibility (Enhanced)
    
    Client->>ProviderRegistry: executeRequest(endpoint: chat/completions URL)
    ProviderRegistry->>AzureAnthropicProvider: executeRequest()
    AzureAnthropicProvider->>AzureAnthropicProvider: Detect chat completions endpoint
    AzureAnthropicProvider->>AzureAnthropicProvider: Extract baseURL, deployment, api-version
    AzureAnthropicProvider->>AzureEndpoint: Use chat completions API
    AzureEndpoint-->>Client: Response

Loading

[greptile-apps]

_{5 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

[waleedlatif1]

@cursor review

[waleedlatif1]

@cursor review

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}