1
0
Fork
You've already forked BlameFree
0
Its an LLM Review Platform built with Rust, think coderabbit but made by an idiot that didnt want to pay 25 cents per file for a code review.
  • Rust 94.7%
  • CSS 4.5%
  • Shell 0.5%
  • Handlebars 0.3%
RacciDev-Bot aa311b114d refactor: fold crb-judge into crb-benchmark, reorganize types 
- Move JudgeVerdict type to crb-types
- Move Metrics type to crb-reporting
- Create crb-benchmark/src/judge.rs with all judge logic
- Remove crb_judge usage from crb-harness
- Remove crb_judge usage from crb-consensus
- Remove build_judge from crb-webui-backend
- Remove stale crb-judge dep from crb-shared
- Remove crb-judge crate from workspace
2026年07月13日 09:55:53 +00:00
.cargo fix: set GCC as default linker in .cargo/config.toml 2026年07月07日 11:39:48 +00:00
crates refactor: fold crb-judge into crb-benchmark, reorganize types 2026年07月13日 09:55:53 +00:00
datasets/golden_comments feat: implement harness-mvp-core — Rust project skeleton with core evaluation loop 2026年07月07日 11:33:00 +00:00
openspec refactor: remove last reminents of once_cell usage 2026年07月13日 06:44:04 +00:00
prompts chore: deslop, hold the slop. 2026年07月08日 07:40:07 +00:00
scripts fix: align Jaccard tokenization to Python whitespace-only split for 100% cross-validation match 2026年07月07日 11:43:42 +00:00
.gitignore fix: missed cache dir due to gitignore pattern overmatching 2026年07月13日 09:52:50 +00:00
Cargo.lock refactor: fold crb-judge into crb-benchmark, reorganize types 2026年07月13日 09:55:53 +00:00
Cargo.toml feat(cache): implement trait-based cache system with derive macro 2026年07月13日 09:55:53 +00:00
LICENSE.md feat: implement harness-mvp-core — Rust project skeleton with core evaluation loop 2026年07月07日 11:33:00 +00:00
mcp_servers.toml feat: agent tool calling — ShellTool, ReadFileTool, GitTool, MCPTool with per-role assignment and call budget 2026年07月07日 11:39:48 +00:00
README.md docs: add README for crb-adapter, audit and fix stale READMEs across workspace 2026年07月07日 11:48:37 +00:00

review-harness

Rust-powered code review harness. Evaluates AI code review agents against git diffs using multi-agent LLM reviewers.

Quick start

# Set up your API key (OpenAI or OpenRouter)
export OPENAI_API_KEY="sk-..."
# For OpenRouter (optional):
export OPENAI_BASE_URL="https://openrouter.ai/api/v1"
# Review working-tree changes
cargo run --release --bin crb-harness -- review --working --model deepseek/deepseek-v4-pro

CLI usage

# Review working-tree changes
cargo run --release --bin crb-harness -- review --working
# Review a specific commit range
cargo run --release --bin crb-harness -- review --commits HEAD~3..HEAD
# Review changes in a specific repo with a custom model
cargo run --release --bin crb-harness -- review \
 --working \
 --path /path/to/repo \
 --model gpt-4o
Flag Env Default Description
--working — false Review working-tree changes (unstaged + staged)
--commits — — Commit range to review (format: base..head)
--path — . Path to the git repository
--model MODEL deepseek/deepseek-v4-pro Model for agent reviews