mirror of
https://github.com/0xMassi/webclaw.git
synced 2026-05-13 17:02:36 +02:00
- Add LlmError::Subprocess(#[from] io::Error) and LlmError::Timeout variants - Implement GeminiCliProvider: new(model) -> Self matching OllamaProvider pattern - Prompts passed exclusively via stdin (Stdio::piped), never as CLI args - 30s subprocess timeout via tokio::time::timeout to prevent hung processes - 6-slot Semaphore to bound concurrent subprocess spawns in MCP context - Stderr captured and included (first 500 bytes) in non-zero exit errors - is_available(): pure `gemini --version` PATH check, no live inference - GEMINI_MODEL env override; default model gemini-2.5-pro - strip_thinking_tags + strip_code_fences applied to stdout output |
||
|---|---|---|
| .. | ||
| src | ||
| Cargo.toml | ||