Plexmcp Oss

🚀 CodeRabbit Review Storage

This directory is designed to store all CodeRabbit code review results. It enables searchable historical analysis through the PlexMCP memory bank, providing valuable insights into past code reviews.

🚀 Quick Start

The CodeRabbit review storage system automatically stores reviews after each commit. To start using it, simply make a commit, and the post - commit hook will handle the rest.

✨ Features

• Historical Analysis: Store and search through past code reviews for learning and auditing purposes.
• Automated Storage: Reviews are automatically stored after each commit.
• Searchable via MCP Tool: Use the find_code_quality_issues() MCP tool to search review history.
• Multiple File Formats: Supports both JSON and Markdown file formats for different use cases.

📦 Installation

There is no specific installation required for the storage system itself. However, ensure that the following are set up correctly:

• CodeRabbit: Install and authenticate CodeRabbit.
• Post - commit Hook: Make sure the post - commit hook in .git/hooks/post - commit is executable.

chmod +x .git/hooks/post-commit

💻 Usage Examples

Basic Usage

After making a commit, the review will be automatically stored in the .coderabbit/reviews/ directory.

Advanced Usage

Custom Review Queries

find_code_quality_issues({
  query: "database query performance optimization",
  severity: "high",
  file_pattern: "crates/api/src/db",
  max_results: 20
})

Trend Analysis

# Count issues per month
for month in {01..12}; do
  count=$(ls .coderabbit/reviews/2025-$month-*.md 2>/dev/null | wc -l)
  echo "2025-$month: $count reviews"
done

Export for External Analysis

# Extract all issues to CSV
node scripts/export-reviews-to-csv.js > reviews.csv

📚 Documentation

Directory Structure

.coderabbit/
├── README.md                          # This file
└── reviews/                           # Review storage
    ├── 2025-12-29-143022-abc1234.json # Raw CodeRabbit output
    ├── 2025-12-29-143022-abc1234.md   # Human-readable summary
    ├── 2025-12-29-151530-def5678.json
    └── 2025-12-29-151530-def5678.md

File Naming Convention

All review files follow this pattern:

{YYYY-MM-DD}-{HHMMSS}-{commit-hash}.{extension}

Example: 2025-12-29-143022-abc1234.json

• Date: 2025 - 12 - 29 (December 29, 2025)
• Time: 14:30:22 (2:30:22 PM UTC)
• Commit: abc1234 (short commit hash)
• Extension: .json or .md

File Formats

JSON Files (.json)

Raw CodeRabbit API response containing structured review data. Schema:

{
  "commit_hash": "abc1234",
  "timestamp": "2025-12-29T14:30:22Z",
  "files_count": 8,
  "issues": [
    {
      "id": "issue-001",
      "severity": "high",
      "category": "security",
      "title": "SQL Injection Vulnerability",
      "description": "User input not sanitized in database query",
      "file": "crates/api/src/auth.rs",
      "line": 45,
      "column": 12,
      "suggestion": "Use parameterized queries with sqlx::query!",
      "code_snippet": "...",
      "impact": "Critical - allows arbitrary SQL execution"
    }
  ],
  "summary": {
    "critical": 2,
    "high": 5,
    "medium": 12,
    "low": 8
  }
}

Markdown Files (.md)

Human - readable summary generated from JSON for embedding indexing. Format:

# CodeRabbit Review - abc1234

**Date:** 2025-12-29T14:30:22Z
**Files Reviewed:** 8
**Issues Found:** 27

## Summary

- Critical: 2
- High: 5
- Medium: 12
- Low: 8

## Issues

### 1. SQL Injection Vulnerability (critical)

**File:** `crates/api/src/auth.rs:45`
**Category:** security

**Description:** User input not sanitized in database query

**Suggested Fix:**
Use parameterized queries with sqlx::query!

How Reviews Are Created

Reviews are automatically stored after each commit via the post - commit hook:

# .git/hooks/post-commit
COMMIT_HASH=$(git rev-parse --short HEAD)
TIMESTAMP=$(date +"%Y-%m-%d-%H%M%S")

# Generate review
coderabbit --prompt-only --type committed --output json \
  > .coderabbit/reviews/$TIMESTAMP-$COMMIT_HASH.json

# Generate markdown summary
node scripts/generate-review-summary.js \
  .coderabbit/reviews/$TIMESTAMP-$COMMIT_HASH.json \
  > .coderabbit/reviews/$TIMESTAMP-$COMMIT_HASH.md

Trigger: Every git commit (bypassed with --no-verify)

Memory Bank Integration

Automatic Indexing

The PlexMCP embeddings watcher automatically indexes markdown review files:

1. File Watcher detects new .md files in .coderabbit/reviews/
2. Chunker splits review into semantic sections
3. Embeddings generated using Ollama (nomic - embed - text)
4. FAISS Store indexes vectors for similarity search
5. SQLite FTS5 indexes text for keyword search Domain: coderabbit-reviews

Searchable via MCP Tool

Claude Code can search review history using the find_code_quality_issues() MCP tool:

// Example tool call
find_code_quality_issues({
  query: "authentication security issues",
  severity: "high",
  file_pattern: "auth.rs",
  max_results: 10
})

Returns:

• Relevant past code reviews
• Issue descriptions and solutions
• File locations and line numbers
• Temporal patterns (recurring issues)

Use Cases

1. Learning from Past Mistakes

Query: "What SQL injection vulnerabilities have we had?" Result: All past reviews mentioning SQL injection, showing:

• Where they occurred
• How they were fixed
• Prevention strategies

2. Tracking Recurring Issues

Query: "Show me all unwrap() issues in the last month" Result: Pattern analysis of .unwrap() usage over time, helping identify:

• Files with repeated issues
• Developers needing training
• Areas needing refactoring

3. Security Audit Trail

Query: "Critical security issues in authentication code" Result: Complete history of security findings in auth - related files:

• What was found
• When it was fixed
• Who committed the fix

4. Onboarding New Developers

Query: "Common code quality issues in this codebase" Result: Aggregated patterns showing:

• Project - specific coding standards
• Common pitfalls to avoid
• Best practices examples

5. Release Quality Checks

Query: "Issues found since last release tag" Result: All quality issues introduced in current release cycle:

• Severity distribution
• File impact analysis
• Regression detection

Maintenance

Storage Size

Growth rate: ~10 - 50 KB per commit (varies by changeset size) Estimated annual storage: ~10 - 50 MB (assuming 1,000 commits/year) Recommendation: Retain indefinitely (low cost, high value)

Cleanup (if needed)

Archive old reviews:

# Archive reviews older than 1 year
find .coderabbit/reviews -name "*.json" -mtime +365 -exec gzip {} \;
find .coderabbit/reviews -name "*.md" -mtime +365 -exec gzip {} \;

Delete very old reviews:

# Delete reviews older than 2 years (CAUTION)
find .coderabbit/reviews -name "*.json" -mtime +730 -delete
find .coderabbit/reviews -name "*.md" -mtime +730 -delete

# Re-index embeddings after deletion
cd .embeddings && npm run reindex

Corruption Recovery

If JSON is corrupted:

1. Review still exists in git history:

git log --all --full-history -- .coderabbit/reviews/
git show <commit-hash>:.coderabbit/reviews/<filename>

2. Re - generate from commit:

git checkout <commit-hash>
coderabbit --type committed --output json > recovered.json

Git Integration

Should Reviews Be Committed?

Yes - Reviews should be committed to git because:

1. Version History - Links reviews to specific commits
2. Team Sharing - All team members access same review history
3. CI/CD Integration - GitHub Actions can analyze trends
4. Backup - Reviews preserved in git history
5. Auditability - Complete quality trail

.gitignore Considerations

Do NOT ignore .coderabbit/reviews/ - these should be committed. Current .gitignore:

# Do NOT ignore CodeRabbit reviews
# !.coderabbit/

# Temporary review files (if any)
.coderabbit/*.tmp
.coderabbit/*.lock

Large Repository Concerns

If review storage becomes too large (unlikely): Option 1: Git LFS for JSON files

git lfs track ".coderabbit/reviews/*.json"

Option 2: Separate review repository

# Create separate repo
git init coderabbit-reviews
git remote add reviews https://github.com/org/plexmcp-reviews

# Store reviews there instead

Troubleshooting

No reviews being created

Check:

# 1. Post-commit hook exists and is executable
ls -la .git/hooks/post-commit
chmod +x .git/hooks/post-commit

# 2. Scripts exist and are executable
ls -la scripts/store-coderabbit-review.sh
chmod +x scripts/store-coderabbit-review.sh

# 3. CodeRabbit authenticated
coderabbit auth status

Manual trigger:

bash scripts/store-coderabbit-review.sh

Reviews not searchable

Check:

# 1. Markdown files exist
ls .coderabbit/reviews/*.md

# 2. Watcher is indexing them
cd .embeddings && npm run stats
# Should show coderabbit-reviews domain

# 3. Re-index if needed
npm run reindex

Reviews are empty

Cause: CodeRabbit execution failed (network, auth, timeout) Fix:

# Check last commit review manually
coderabbit --type committed --output json

# If works, re-generate review
bash scripts/store-coderabbit-review.sh

Best Practices

1. Commit Reviews Immediately

Reviews are most valuable when committed alongside the code they review. Workflow:

git add .
git commit -m "Implement feature X"
# Post-commit hook runs automatically, stores review

git add .coderabbit/reviews/
git commit -m "Add CodeRabbit review for feature X"
git push

Alternative: Include review in same commit (requires custom hook)

2. Review Before Major Releases

Before releases, search for unresolved issues:

// Via Claude Code
find_code_quality_issues({
  query: "critical high severity issues",
  severity: "critical"
})

3. Periodic Quality Audits

Monthly review of patterns:

# Generate summary report
cd .coderabbit/reviews
grep -h "^### " *.md | sort | uniq -c | sort -rn

4. Link Reviews to PRs

In PR descriptions, reference review files:

## Code Quality

CodeRabbit review: `.coderabbit/reviews/2025-12-29-143022-abc1234.md`

**Issues found**: 5 (3 high, 2 medium)
**All resolved**: ✅

Last Updated: 2025 - 12 - 29 Maintained By: PlexMCP Development Team