Release v0.10.0

Gemini Support - Google Gemini API Provider with Streaming

Adds Google Gemini as the fourth LLM provider, with full streaming support, safety content filtering, and CI mode integration.

Added

Google Gemini Provider

• New GeminiProvider implementing the ApiBackend trait
• Supports generateContent and streamGenerateContent (SSE) endpoints
• Uses x-goog-api-key header authentication
• Default model: gemini-3-flash-preview
• Default endpoint: https://generativelanguage.googleapis.com
• Configurable max_tokens, temperature, and custom endpoint

Gemini Streaming Support

• Full SSE streaming with real-time typing effect
• Handles finishReason variants: STOP, MAX_TOKENS, SAFETY, RECITATION, etc.
• Non-STOP finish reasons display localized warnings
• Parse error counting and reporting (consistent with Claude/OpenAI streaming)

Gemini Safety Content Filtering

• Detects and reports content blocked by Gemini safety filters
• Returns clear error messages with the blocking reason (e.g., SAFETY)

Gemini API Validation

• validate() sends minimal test request (max_output_tokens=1) to verify connectivity
• Works with gcop config validate command

CI Mode Gemini Support

• GCOP_CI_PROVIDER=gemini now supported
• CI mode provider list updated: claude, openai, ollama, gemini

Gemini Error Suggestions

• API key errors now suggest: Add 'api_key = "AIza..."' to [llm.providers.gemini]
• Provider-not-found suggestions updated to include gemini

Changed

ApiStyle Enum Extended

• New Gemini variant with Display, FromStr, and serde support
• default_model() returns "gemini-3-flash-preview" for Gemini

i18n Messages Updated

• Added Gemini-specific messages in both English and Chinese locales
• Updated CI provider error messages to include "gemini"
• Updated provider-not-found suggestions to include "gemini"

Configuration

[llm]
default_provider = "gemini"

[llm.providers.gemini]
api_key = "AIza..."
model = "gemini-3-flash-preview"        # optional
endpoint = "https://generativelanguage.googleapis.com"  # optional
max_tokens = 2000                        # optional
temperature = 0.3                        # optional

Road to 1.0

v0.10.0 completes the four-provider lineup (Claude, OpenAI, Ollama, Gemini). If no further breaking changes are needed, the next minor release will be v1.0.0, marking API stability.

Upgrade

# Homebrew
brew upgrade gcop-rs

# Cargo
cargo install gcop-rs

# pip
pip install --upgrade gcop-rs