@kb-labs/review-cli

CLI commands for AI Review plugin.

Commands

review:run

Run code review analysis.

# Heuristic mode (CI, fast, no LLM)
kb review run

# Full mode (heuristic + LLM)
kb review run --mode=full

# LLM-only mode (deep analysis)
kb review run --mode=llm

# Scope options
kb review run --scope=all       # Entire codebase
kb review run --scope=changed   # Changed files vs main (default)
kb review run --scope=staged    # Staged files only

# Use preset
kb review run --preset=typescript-strict

# Custom file patterns
kb review run --files="src/**/*.ts" --files="src/**/*.tsx"

# JSON output
kb review run --json

Review Modes

heuristic (CI Mode)

Fast, deterministic analysis using only heuristic engines:

• ESLint (TypeScript/JavaScript)
• Ruff (Python) - coming soon
• golangci-lint (Go) - coming soon
• Clippy (Rust) - coming soon

Use cases:

• CI/CD pipelines
• Pre-commit hooks
• Quick local checks

Exit code:

• 0: No blocker or high severity issues
• 1: Has blocker or high severity issues

full (Local Mode)

Comprehensive analysis with heuristic + LLM:

• All heuristic engines
• LLM analyzers for complex issues
• Cached LLM results

Use cases:

• Local development
• Pre-PR review
• Weekly codebase audits

Exit code:

• 0: No blocker issues
• 1: Has blocker issues

llm (Deep Analysis Mode)

LLM-only analysis for semantic issues:

• Naming conventions
• Architecture patterns
• Logic bugs
• Code smells

Use cases:

• Architecture reviews
• Complex refactoring
• Semantic bug detection

Exit code:

• 0: No blocker issues
• 1: Has blocker issues

Scope Options

all

Analyze entire codebase:

kb review run --scope=all

changed

Analyze files changed vs main branch (git diff):

kb review run --scope=changed  # default

staged

Analyze only staged files (git diff --staged):

kb review run --scope=staged

Presets

Use predefined rule sets:

# TypeScript strict preset
kb review run --preset=typescript-strict

# Python production preset
kb review run --preset=python-production

# Security-focused preset
kb review run --preset=security-all

Output Formats

Text (default)

Human-readable output with colors:

Running heuristic analysis...
✔ Found 3 issue(s)

┌─ Code Review ─────────────────┐
│ Summary                       │
│ Files: 5                      │
│ Findings: 3                   │
│ High: 1                       │
│ Medium: 2                     │
│ Engines: eslint               │
│                               │
│ Findings                      │
│ [HIGH] src/index.ts:10 - ...  │
│ [MEDIUM] src/utils.ts:5 - ... │
│ [MEDIUM] src/api.ts:20 - ...  │
└───────────────────────────────┘

JSON

Machine-readable output:

kb review run --json

{
  "findings": [
    {
      "id": "eslint:src/index.ts:no-unused-vars:10:5",
      "ruleId": "no-unused-vars",
      "type": "code-quality",
      "severity": "medium",
      "confidence": "certain",
      "file": "src/index.ts",
      "line": 10,
      "message": "'foo' is defined but never used.",
      "engine": "eslint",
      "source": "heuristic"
    }
  ],
  "metadata": {
    "mode": "heuristic",
    "scope": "changed",
    "analyzedFiles": 5,
    "duration": 1234,
    "engines": ["eslint"],
    "timestamp": "2025-01-21T12:00:00Z"
  }
}

Examples

CI Pipeline

# .github/workflows/review.yml
- name: Code Review
  run: |
    kb review run --mode=heuristic --scope=changed --json > review.json

    # Fail if issues found
    if [ $? -ne 0 ]; then
      echo "Code review failed"
      exit 1
    fi

Pre-commit Hook

#!/bin/bash
# .git/hooks/pre-commit

kb review run --mode=heuristic --scope=staged

if [ $? -ne 0 ]; then
  echo "Fix issues before committing"
  exit 1
fi

Local Development

# Quick check
kb review run

# Comprehensive review before PR
kb review run --mode=full --scope=all

# Deep analysis for refactoring
kb review run --mode=llm --files="src/core/**/*.ts"

Configuration

Configure via .kb/kb.config.json:

{
  "plugins": {
    "@kb-labs/review": {
      "eslintConfig": ".eslintrc.json",
      "defaultMode": "heuristic",
      "defaultScope": "changed"
    }
  }
}

Platform Integration

The CLI uses KB Labs platform composables:

import { useLLM, useCache, useAnalytics, useLoader, useConfig } from '@kb-labs/sdk';

// Access LLM (if configured)
const llm = useLLM({ tier: 'medium' });

// Access cache
const cache = useCache();

// Show progress
const loader = useLoader('Analyzing...');
loader.start();

Exit Codes

• 0: Success (no critical issues)
• 1: Issues found (blockers or high severity in heuristic mode)
• 1: Error during execution

Architecture

review:run
│
├─ ReviewOrchestrator
│   ├─ runHeuristicAnalysis()
│   │   ├─ ESLint adapter
│   │   ├─ Ruff adapter (TODO)
│   │   └─ deduplicateFindings()
│   │
│   ├─ runFullAnalysis()
│   │   ├─ heuristic + LLM
│   │   └─ deduplicateFindings()
│   │
│   └─ runLLMAnalysis()
│       └─ LLM-only (TODO)
│
└─ Output (text/json)

Related Packages

• @kb-labs/review-contracts - Type definitions
• @kb-labs/review-heuristic - Heuristic engines (ESLint, etc.)
• @kb-labs/review-core - Orchestration logic