🔬 SmartBench v0.6.0
LLM-powered universal code diagnostic platform — 14 languages, multi-agent debate, RAG + evidence verification. Zero hallucination guarantee.
✨ Highlights
- 14 languages + 20+ frameworks supported
- 5-stage diagnostic pipeline
- Multi-agent debate engine with evidence verification (Proposer → Verifier → Critique → Judge)
- RAG vector retrieval with 3-tier embedding fallback
- Code graph engine (12-language AST parsing)
- 5 verified strategy templates
- 8 LLM providers auto-detected
- API keys in-memory only, never on disk
📦 Installation
pip install smartbench-0.6.0-py3-none-any.whl
# with RAG support:
pip install smartbench-0.6.0-py3-none-any.whl[rag]
# or from GitHub:
pip install git+https://github.com/xianyu-sheng/SmartBench.git