seraph-ai 0.2.1

Code quality gate — mutation testing, static analysis, and risk scoring for AI-generated code

Seraph

Code quality gate for AI-generated code. Mutation testing, static analysis, flaky test detection, and risk scoring in one pipeline.

Seraph takes a git diff and stress-tests it across 5 dimensions before it ships.

Install

pip install seraph-ai

With optional tools:

pip install "seraph-ai[mutation]"   # mutmut for mutation testing
pip install "seraph-ai[sentinel]"   # git-sentinel for risk signals
pip install "seraph-ai[all]"        # everything

Quick Start

# Assess current changes (skipping heavy steps for speed)
seraph assess --skip-baseline --skip-mutations

# Full assessment with mutation testing
seraph assess

# View past assessments
seraph history

# JSON output for CI integration
seraph assess --json

# Debug mode
seraph assess --verbose

What It Does

Seraph runs your code changes through 5 tests and produces an overall grade (A-F):

Dimension	Weight	What It Checks
Mutation Score	30%	Deliberately breaks your code — do your tests catch it?
Static Cleanliness	20%	Ruff + mypy findings weighted by severity
Test Baseline	15%	Runs tests 3x to detect flaky tests
Sentinel Risk	20%	Hot files, pitfall patterns, and risk signals from git history
Co-change Coverage	15%	Did you forget files that historically change together?

Grades: A >= 90, B >= 75, C >= 60, D >= 40, F < 40

MCP Server

Seraph works as an MCP server for AI assistants (Claude, etc.):

seraph-mcp

Tools: seraph_assess, seraph_mutate, seraph_history, seraph_feedback

Claude Desktop Config

{
  "mcpServers": {
    "seraph": {
      "command": "seraph-mcp",
      "env": {
        "SERAPH_REPO_PATH": "/path/to/your/repo"
      }
    }
  }
}

Configuration

Create .seraph/config.toml in your repo to customize:

[timeouts]
mutation_per_file = 300
static_analysis = 60

[scoring]
mutation_weight = 0.40
static_weight = 0.15

[pipeline]
baseline_runs = 5

[retention]
retention_days = 30

[logging]
level = "DEBUG"

Environment variables override TOML (e.g. SERAPH_TIMEOUT_MUTATION_PER_FILE=300).

Commands

seraph assess [repo_path]    # Run full assessment
seraph history [repo_path]   # View past assessments
seraph feedback <id> <outcome>  # Submit feedback (accepted/rejected/modified)
seraph prune [repo_path]     # Delete old data (--days N --yes)

Works With

• Sentinel — Git history intelligence (hot files, co-change patterns, pitfall detection)
• mutmut — Mutation testing
• ruff — Fast Python linter
• mypy — Type checker

Part of the EvoIntel MCP Suite

Seraph solves AI Blindness #4: Code Quality — mutation survival, flaky tests, and risk signals that "all tests pass" will never reveal.

Part of the EvoIntel MCP Suite by Evolving Intelligence AI: five tools for five blindnesses no model improvement will ever fix.

Tool	Blindness	Install
Sentinel	Project History	pip install git-sentinel
Niobe	Runtime Behavior	pip install niobe
Merovingian	Cross-Service Dependencies	pip install merovingian
Seraph	Code Quality	pip install seraph-ai
Anno	Web Content	npm install -g @evointel/anno

License

MIT