Release v0.5.0

New Features

Provider Fallback Support

gcop-rs now supports automatic failover between LLM providers. When your primary provider fails (network error, rate limit, etc.), it will automatically try the next provider in your fallback list.

Configuration:

[llm]
default_provider = "claude"
fallback_providers = ["openai", "ollama"]

Behavior:

• Tries primary provider first
• On failure, shows warning and switches to next provider
• Works for both commit message generation and code review
• Supports streaming mode (falls back to non-streaming if all streaming providers fail)

Claude Streaming Support

Claude provider now supports streaming responses with real-time typing effect, just like ChatGPT.

• Uses SSE (Server-Sent Events) for efficient streaming
• Shows characters as they're generated
• Graceful fallback to spinner mode if streaming fails

Enhanced Retry Mechanism

The API retry system now supports:

• Retry-After header: Respects rate limit hints from APIs (HTTP 429 responses)
• Configurable max delay: New max_retry_delay_ms option to cap wait time

[network]
max_retries = 3
retry_delay_ms = 1000
max_retry_delay_ms = 60000  # Maximum 60 seconds

Colored Provider Output

LLM providers now display colored warning and info messages when:

• Switching to fallback provider
• Retrying after rate limit
• Encountering recoverable errors

Improvements

Better User Experience

• Improved commit command interaction flow
• Better review command output formatting
• Clearer progress indicators

Error Handling

• Restructured error types for clarity
• More user-friendly error messages
• Better suggestions for common issues

Code Quality

• Extracted common prompt building logic
• Unified response processing across providers
• Reduced code duplication in LLM module

Bug Fixes

• Streaming Error Handling: Fixed error handling and log levels for streaming responses

Documentation

• Updated streaming output documentation
• Added Claude configuration examples
• Added installation update/uninstall instructions for Homebrew, pipx, cargo-binstall, and cargo

Upgrade Notes

Upgrading from v0.4.x requires no action, fully backward compatible.

Optional new configuration options:

[llm]
fallback_providers = ["openai", "ollama"]  # Optional failover list

[network]
max_retry_delay_ms = 60000  # Optional max wait time

Installation

# Homebrew (macOS/Linux)
brew tap AptS-1547/gcop-rs
brew install gcop-rs

# pipx (Python users, recommended)
pipx install gcop-rs

# cargo-binstall (no compilation)
cargo binstall gcop-rs

# cargo install (compile from source)
cargo install gcop-rs

Or download pre-built binaries from Releases.

Feedback

If you have any issues or suggestions, please submit an Issue.