docs: add CLAUDE.md for project build and development guidance

Author: hanrw

CLAUDE.md

@@ -0,0 +1,104 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Build and Development Commands
+
+### Building the Project
+```bash
+./gradlew build                 # Build all modules
+./gradlew clean build           # Clean and build all modules
+./gradlew assemble             # Assemble outputs without running tests
+```
+
+### Running Tests
+```bash
+./gradlew test                  # Run tests for all platforms
+./gradlew jvmTest              # Run JVM tests only
+./gradlew macosArm64Test      # Run macOS ARM64 tests
+./gradlew iosSimulatorArm64Test # Run iOS simulator tests
+./gradlew allTests             # Run tests for all targets with aggregated report
+./gradlew check                # Run all verification tasks
+```
+
+### Test Coverage
+```bash
+./gradlew koverHtmlReport      # Generate HTML coverage report for all code
+./gradlew koverXmlReport       # Generate XML coverage report
+./gradlew koverVerify          # Run coverage verification (min 86% required)
+```
+
+### Running Specific Module Tests
+```bash
+./gradlew :openai-client:openai-client-core:test
+./gradlew :anthropic-client:anthropic-client-core:test
+./gradlew :ollama-client:ollama-client-core:test
+./gradlew :gemini-client:gemini-client-core:test
+./gradlew :openai-gateway:openai-gateway-core:test
+```
+
+## Project Architecture
+
+This is a Kotlin Multiplatform project providing AI/LLM client implementations for multiple providers. The codebase follows a modular architecture with clear separation of concerns.
+
+### Core Architecture Patterns
+
+1. **Multiplatform Structure**: Each client module has platform-specific implementations:
+   - `-core`: Common implementation shared across platforms
+   - `-darwin`: Apple platform specific implementations (iOS, macOS)
+   - `-cio`: JVM-specific implementations using CIO
+
+2. **Provider Pattern**: The `openai-gateway` module implements a gateway pattern that allows switching between different LLM providers (OpenAI, Anthropic, Ollama, Gemini) using a unified interface.
+
+3. **HTTP Client Abstraction**: The `common` module provides a shared `HttpRequester` interface that abstracts HTTP operations across platforms, using Ktor under the hood.
+
+4. **Dependency Injection**: Uses Koin for dependency injection across the codebase, with platform-specific configurations.
+
+### Module Structure
+
+- **common/**: Shared networking and utility code
+  - HTTP client abstraction (`HttpRequester`)
+  - Ktor configuration for different platforms
+  - JSON serialization utilities
+
+- **openai-client/**: OpenAI API client implementation
+  - Chat completions, images, and legacy completions APIs
+  - Streaming support for chat completions
+
+- **anthropic-client/**: Anthropic Claude API client
+  - Messages API implementation
+  - Image support with base64 encoding
+
+- **ollama-client/**: Ollama local LLM client
+  - Chat and generate endpoints
+  - Local model management
+
+- **gemini-client/**: Google Gemini API client
+  - Text generation with multimodal support
+
+- **openai-gateway/**: Unified gateway for all providers
+  - Provider abstraction allowing runtime switching
+  - Adapter pattern to convert between provider-specific and OpenAI formats
+  - Extensions for converting requests/responses between different provider formats
+
+### Key Interfaces
+
+- `OpenAI`: Main interface for OpenAI operations (Chat, Images, Completions)
+- `OpenAIGateway`: Gateway interface for multi-provider support
+- `HttpRequester`: HTTP client abstraction for cross-platform requests
+- `OpenAIProvider`: Provider interface for different LLM services
+
+### Configuration
+
+Each client uses a configuration pattern (e.g., `OpenAIConfig`, `AnthropicConfig`) that requires:
+- Base URL (with defaults for each provider)
+- API key
+- Optional provider-specific settings
+
+### Testing Strategy
+
+The project uses:
+- Unit tests with mocked HTTP clients (`MockHttpClient`)
+- Integration tests (files ending with `ITest`) for actual API calls
+- Platform-specific test configurations
+- Kover for code coverage with 86% minimum threshold