TetherAI v0.2.0

�� New Providers

Grok AI Provider (@tetherai/grok@0.1.0)

• xAI Integration: Native support for Grok models
• Models: grok-beta, grok-beta-vision, grok-beta-2, grok-beta-2-vision
• Features: Streaming chat, non-streaming chat, model management
• Context: Up to 128K tokens for vision models

Local LLM Provider (@tetherai/local@0.1.0)

• Local Endpoints: Ollama, LM Studio, custom APIs
• Flexible Models: Any OpenAI-compatible model
• Features: Streaming chat, model validation, token limits
• Use Cases: Local development, offline AI, custom deployments

🔄 Enhanced Existing Providers

OpenAI Provider (@tetherai/openai@0.2.0)

• Bug Fixes: Usage data parsing (prompt_tokens → promptTokens)
• Improvements: Better error handling and type safety

Anthropic Provider (@tetherai/anthropic@0.2.0)

• Bug Fixes: Usage data parsing (input_tokens → promptTokens)
• Improvements: Enhanced streaming response handling

Mistral Provider (@tetherai/mistral@0.2.0)

• Bug Fixes: Type definitions and error handling
• Improvements: Better API response parsing

Enhanced Chat Options & Model Management

Added

• Enhanced chat options (temperature, maxTokens, topP, etc.)
• Model management (getModels, validateModel, getMaxTokens)
• Non-streaming chat method
• Middleware compatibility improvements

Updated

• OpenAI provider to v0.1.0
• Anthropic provider to v0.1.0
• Comprehensive documentation
• Test coverage

Anthropic Provider v0.0.1, OpenAI Provider v0.0.4

openai

• Refactored build output to dist/src/*
• Added stricter TypeScript configs
• Verified working in Next.js example
• CI extended with artifact checks

anthropic

• Initial release
• Streaming-first Anthropic adapter
• Retry & fallback middleware included

First release of TetherAI with OpenAI provider (@tetherai/openai)

– Streaming-first chat API
– Retry and fallback middleware
– Next.js example app