feat: add AI Stupid Level smart router integration

[StudioPlatforms]

Overview

Adds support for AI Stupid Level's intelligent model routing as an alternative LLM provider alongside Anthropic.

Features

• Multi-provider architecture (Anthropic + AI Stupid Level)
• 6 auto-routing models with intelligent selection
• Environment-based configuration
• Full backward compatibility
• TypeScript errors fixed

Configuration

# Use AI Stupid Level
LLM_PROVIDER=aistupidlevel
AI_STUPID_LEVEL_API_KEY=your_key_from_https://aistupidlevel.info/router/keys
AI_STUPID_LEVEL_MODEL=auto-coding  # optional, defaults to auto-coding

Available Models

• auto - Best overall model based on real-time benchmarks
• auto-coding - Optimized for coding tasks (default)
• auto-reasoning - Complex reasoning tasks
• auto-creative - Creative writing
• auto-fastest - Prioritizes speed
• auto-cheapest - Cost optimization

Changes

• Added @ai-sdk/openai package for OpenAI-compatible API support
• Implemented getAIStupidLevelModel() function with custom baseURL
• Added provider selection via LLM_PROVIDER environment variable
• Added TypeScript types for new environment variables
• Fixed pre-existing TypeScript errors in stream-text.ts
• Maintains full backward compatibility with existing Anthropic integration

Testing

• ✅ TypeScript compilation (0 errors)
• ✅ Backward compatibility maintained
• ✅ Provider routing verified

[Nephew209-oss]

I don't believe that action with that certain language is appropriate in my opinion but I wouldn't approve these changes or that call

[StudioPlatforms]

Thank you for reviewing this PR. I understand the name 'AI Stupid Level' might seem unconventional, but I want to clarify the intent behind it.

The name is intentionally ironic and self-aware - it's a playful reference to the common frustration when AI models produce incorrect responses. Rather than being disrespectful, it's a humble acknowledgment that no AI is perfect, and our platform helps users intelligently route between models to minimize those moments.

Despite the lighthearted name, this is serious production-grade technology with intelligent model selection, real-time benchmarks, and professional-grade reliability. The tech industry has a long tradition of playful names for serious projects (Git, Apache, etc.).

I'd encourage evaluating this PR on its technical merit: clean code, full backward compatibility, zero breaking changes, proper TypeScript typing, and comprehensive testing. The integration is optional and doesn't affect existing Anthropic users.

I respect that naming preferences are subjective, but there's no malicious intent here. If you have specific technical concerns about the implementation, I'm happy to address them.

[igun997]

hey https://aistupidlevel.info it's a platform brooooo. that a brand not just a word "stupid"

I don't believe that action with that certain language is appropriate in my opinion but I wouldn't approve these changes or that call