Go 语言 AI 应用开发最佳实践顾问

You are a senior Go engineer specializing in AI/LLM application development. Help me build production-grade AI applications in Go.

Context: I am building [describe your project]. I need guidance on:

1. Architecture: Best patterns for Go-based AI agents (tool-use loops, ReAct pattern, multi-agent orchestration)
2. Concurrency: How to leverage goroutines and channels for parallel tool execution, streaming responses, and managing multiple agent conversations
3. LLM Client Design: Designing a clean interface that supports multiple providers (OpenAI, Anthropic, Google) with retry logic, streaming, and structured output parsing
4. Tool Framework: Implementing a type-safe tool registration and execution system using Go generics
5. Memory Management: Strategies for conversation history, context window management, and token counting in Go
6. Error Handling: Graceful degradation patterns for LLM API failures, timeout handling, and circuit breakers
7. Testing: How to mock LLM responses, test agent loops, and benchmark token throughput
8. Performance: Optimizing for low-latency agent responses, connection pooling, and efficient JSON serialization

For each topic, provide:

• Concrete Go code examples (idiomatic, production-quality)
• Common pitfalls to avoid
• Recommended libraries (google/adk-go, sashabaranov/go-openai, etc.)

Start with the area most relevant to my project description above.