Release v0.9.1

Code Quality - ProgressReporter Decoupling + ApiStyle Type Safety + Enhanced Config Validation

An internal quality release focused on architectural decoupling, type safety improvements, and configuration validation, with no user-facing feature changes.

Added

ProgressReporter Trait

• New ProgressReporter trait in llm/mod.rs decouples LLM layer from UI layer
• LLM providers no longer depend on ui::Spinner directly
• Enables easier testing and future alternative progress reporters (logging, event bus, etc.)

ApiStyle Enum

• New type-safe ApiStyle enum (Claude, OpenAI, Ollama) replaces string-based API style
• Compile-time exhaustive matching ensures all API styles are handled
• Built-in default_model() method centralizes default model mapping
• Implements FromStr and serde traits for seamless config file compatibility

Configuration Reference Validation

• default_provider is now validated against defined [llm.providers] entries at startup
• fallback_providers entries are validated similarly
• Misconfigured references now fail fast with clear error messages instead of runtime crashes

Machine-Readable Markdown Output

• New OutputFormat::is_machine_readable() method unifies JSON and Markdown behavior
• Markdown output now skips spinner and step indicators (consistent with JSON mode)
• New effective_colored() method auto-disables color for machine-readable formats

Changed

Error Handling Responsibilities Shifted

• review and stats commands now handle their own JSON error output internally
• main.rs no longer needs to know each command's output type
• Extracted handle_command_error() helper to eliminate duplicate error display logic

CI Mode Uses ApiStyle Enum

• CI provider type validation uses ApiStyle::from_str() instead of string matching
• Default model selection uses ApiStyle::default_model() instead of scattered match arms
• Eliminates unreachable!() branches

Configuration Examples Simplified

• Example config files reduced from 144 lines to 39 lines
• Removed verbose comments and advanced options
• Added documentation link for full configuration reference

LLM Provider Interface Updated

• generate_commit_message() and review_code() now accept Option<&dyn ProgressReporter> instead of Option<&Spinner>
• All built-in providers (Claude, OpenAI, Ollama, Fallback) updated accordingly

Upgrade

# Homebrew
brew upgrade gcop-rs

# Cargo
cargo install gcop-rs

# pip
pip install --upgrade gcop-rs