astonai

Aston AI

Aston is a code intelligence system for parsing, analyzing, and finding test coverage gaps in your code.

Latest: v0.3.7 has astonrank (go) to aid criticality scoring

# Quick Setup
pip install astonai
aston init --offline
aston graph build
aston cache warm-up  # Enable high-performance caching

# Key Features
aston coverage --critical-path        # Find high-impact code paths
aston test-suggest core/auth.py --k 3 # Generate test suggestions
aston criticality analyze --top 10    # Identify critical components
aston regression-guard --since HEAD~1 # Analyze change risk

Feature Overview

Version	Feature	Key Benefits	CLI Commands
v0.3.8	Local Embeddings	• Local MiniLM embedding generation • FAISS vector similarity search • No external API dependencies • Repository-centric storage	aston embed --backend minilm aston embed --dry-run aston embed --file src/main.py
v0.3.4	Micro Cache	• Sub-300ms graph operations • 256MB memory limit • Performance monitoring	aston cache warm-up aston cache status aston cache clear
v0.3.3	Criticality Scorer	• Advanced risk assessment • Component importance ranking • Test prioritization	aston criticality analyze aston criticality export aston criticality tune
v0.3.2	Regression Guard	• Multi-factor risk scoring • Threshold-based blocking • CI/CD integration	aston regression-guard --since <ref> aston regression-guard --exit-code
v0.2.5	Test Suggestions	• Intelligent pytest generation • LLM fallback mode • Rich context for developers	aston test-suggest <file> --k 3 aston test-suggest <file> --prompt aston test-suggest <file> --llm
v0.2.3	Filter Engine	• Incremental rechunk • Smart change detection • Pattern-based filtering	aston init --rechunk aston refresh aston init --preset python-only
v0.2.0	Critical Path	• High-impact node detection • Focus testing efforts • Weight-based scoring	aston coverage --critical-path aston coverage --weight loc
v0.1.19	Graph Visualization	• DOT format export • Interactive visualization • Relationship exploration	aston graph export aston graph view
v0.1.18	Relationship Analysis	• Function call tracking • Import dependency mapping • Graph statistics	aston graph build aston graph stats
v0.1.16	Core CLI	• Repository initialization • Coverage gap detection • Test execution	aston init aston coverage aston test

Explain

aston init → creates chunks + nodes aston graph build → uses those to create edges

Installation

# Base installation (fast, minimal dependencies)
pip install astonai

# For database features (Neo4j graph storage)
pip install "astonai[db]"

# For visualization features (matplotlib plotting)
pip install "astonai[viz]"

# For LLM-powered features (OpenAI integration)
pip install "astonai[llm]"

# For local embedding features (MiniLM + FAISS)
pip install "astonai[embed]"

# Everything bundle
pip install "astonai[all]"

# Quiet installation (suppress download chatter)
pip install -q astonai

Installation Notes:

• Base install: Core functionality with offline mode (10 dependencies, ~20-25 packages)
• Database extra: Adds Neo4j for live graph storage when not using --offline mode
• Visualization extra: Adds matplotlib for graph plotting and visualization features
• LLM extra: Adds OpenAI client for advanced test suggestion capabilities
• Embed extra: Adds sentence-transformers and FAISS for local embedding generation
• Version checking: Run aston --version to verify installation

Quick Start

# Initialize your repository
aston init --offline

# Generate knowledge graph relationships
aston graph build

# View knowledge graph statistics
aston graph stats

# Smart incremental updates (recommended for ongoing development)
aston refresh

# Enable high-performance caching
aston cache warm-up
aston cache status

# Find critical paths and generate test suggestions
aston coverage --critical-path
aston test-suggest core/auth.py --k 3
aston test-suggest user/models.py --prompt --yaml context.yaml

# Analyze code criticality and regression risk
aston criticality analyze --top 10
aston regression-guard --since HEAD~1

Core Commands

Repository Initialization

# Initialize repository and create knowledge graph
aston init [--offline] [--preset PRESET] [--include PATTERN] [--exclude PATTERN]

# Incremental rechunk - fast updates for changed files only
aston init --rechunk [--offline]

# Force full rebuild
aston init --force [--offline]

Advanced Filtering Options:

• --preset: Apply preset configurations (python-only, no-tests, source-only, minimal)
• --include, -i: Include only files matching these glob patterns (can be used multiple times)
• --exclude, -e: Exclude files matching these glob patterns in addition to defaults (can be used multiple times)
• --include-regex: Include only files matching these regex patterns (can be used multiple times)
• --exclude-regex: Exclude files matching these regex patterns (can be used multiple times)
• --dry-run: Show which files would be processed without actually processing them
• --show-patterns: Display all active filter patterns and exit
• --create-astonignore: Create a template .astonignore file for persistent filtering

Incremental Updates:

• --rechunk: Process only files that have changed since last run (fast incremental updates)
• --force: Force complete rebuild even if knowledge graph exists

Default Excludes: Common directories are automatically excluded:

• venv*/**, .venv*/**, env/**, .env/**
• node_modules/**, .git/**, .svn/**, .hg/**
• __pycache__/**, *.pyc, .pytest_cache/**
• build/**, dist/**, *.egg-info/**
• .idea/**, .vscode/**, and more

Examples:

# Use preset configurations
aston init --preset python-only --offline
aston init --preset no-tests --offline

# Incremental rechunk for fast updates
aston init --rechunk --offline

# Custom filtering with patterns
aston init --include "src/**/*.py" --include "lib/**/*.py" --offline
aston init --exclude "legacy/**" --exclude "deprecated/**" --offline

# Use regex patterns for advanced filtering
aston init --include-regex ".*/(core|utils)/.*\.py$" --offline

# Preview what would be processed
aston init --preset minimal --dry-run

# Create .astonignore template for persistent rules
aston init --create-astonignore

Environment Variables:

• ASTON_INCLUDE_PATTERNS: Comma-separated include patterns
• ASTON_EXCLUDE_PATTERNS: Comma-separated exclude patterns

Intelligent Refresh

# Smart incremental updates with change analysis
aston refresh [--strategy auto|incremental|full] [--force-full] [--dry-run]

The refresh command provides intelligent updates:

• Auto Strategy: Automatically chooses between incremental and full refresh based on changes
• Change Detection: Uses file hashes to detect actual modifications
• Dry Run: Preview what would be updated without making changes
• Force Full: Override auto-detection and force complete rebuild

Examples:

# Smart refresh (recommended)
aston refresh

# Preview changes without applying
aston refresh --dry-run

# Force full refresh
aston refresh --force-full

# Use specific strategy
aston refresh --strategy incremental

Test Coverage

# Run tests with coverage
aston test

# Find testing gaps
aston coverage [--threshold 80] [--json results.json] [--exit-on-gap]

# Identify critical implementation nodes
aston coverage --critical-path [--n 50] [--weight loc]

Options:

• --threshold: Minimum coverage percentage (default: 0)
• --json: Output results in JSON format
• --exit-on-gap: Return code 1 if gaps found (useful for CI)
• --coverage-file: Specify custom coverage file location
• --critical-path: Identify critical code paths that need testing
• --n: Number of critical nodes to return (default: 50)
• --weight: Weight function for critical path (loc, calls, churn)

Knowledge Graph

# Build edge relationships between nodes with advanced filtering
aston graph build [--preset PRESET] [--include PATTERN] [--exclude PATTERN]

# View statistics about the knowledge graph
aston graph stats

# Export graph to DOT format
aston graph export [--output graph.dot] [--filter CALLS,IMPORTS] [--open]

# Open interactive graph viewer in browser
aston graph view [--filter CALLS,IMPORTS]

Advanced Filtering for Graph Build:

• --preset: Apply preset configurations (python-only, no-tests, source-only, minimal)
• --include, -i: Include only files matching these glob patterns (can be used multiple times)
• --exclude, -e: Exclude files matching these glob patterns in addition to defaults (can be used multiple times)
• --include-regex: Include only files matching these regex patterns (can be used multiple times)
• --exclude-regex: Exclude files matching these regex patterns (can be used multiple times)
• --dry-run: Show which files would be processed without actually processing them
• --show-patterns: Display all active filter patterns and exit

Examples:

# Build with preset filtering
aston graph build --preset no-tests

# Include only specific directories
aston graph build --include "src/**/*.py" --include "lib/**/*.py"

# Use regex patterns for advanced filtering
aston graph build --include-regex ".*/(core|utils)/.*\.py$"

# Preview what would be processed
aston graph build --preset python-only --dry-run

The graph command provides:

• build: Analyzes your codebase to extract CALLS and IMPORTS edges with advanced filtering
• stats: Displays node and edge statistics
• export: Exports to Graphviz DOT format for external visualization
• view: Opens interactive D3-force based viewer in browser

Test Suggestions

# Generate test suggestions for a file or function
aston test-suggest <file_or_node> [--k 5] [--llm] [--model gpt-4o]

# Generate rich context for developers or AI agents
aston test-suggest <file_or_node> --prompt [--json context.json]

# Output in multiple formats
aston test-suggest core/auth.py --yaml tests.yaml --json tests.json

# Use LLM with budget control
aston test-suggest api/endpoints.py --llm --budget 0.01 --model gpt-4o

# Debug path resolution issues
aston test-suggest <file_or_node> --debug

# Prioritize tests based on criticality scores
aston test-suggest complex/algorithm.py --k 3 --criticality-ranked

Intelligent Test Generation:

• Heuristic Mode: Lightning-fast pytest skeleton generation (≤1.2s)
• Boundary Value Testing: Automatic edge cases for int/float, string, list, dict, bool types
• Happy Path Coverage: Valid input test cases for comprehensive coverage
• Error Condition Testing: Invalid input handling and exception testing

LLM Integration (Optional):

• Fallback Strategy: Uses LLM when heuristics can't generate suggestions
• Cost Control: Built-in budget tracking and enforcement
• Model Selection: Support for GPT-4o, GPT-4-turbo, GPT-3.5-turbo
• Performance Guarantee: ≤6s latency for LLM-generated suggestions

Rich Context Mode:

• Developer Guidance: Comprehensive test implementation guides
• Parameter Analysis: Detailed function signature and dependency analysis
• Best Practices: pytest patterns and testing recommendations
• AI-Agent Ready: Structured context for automated test generation

Options:

• --k: Number of suggestions to generate (default: 5)
• --llm: Use LLM fallback if heuristics fail (requires OPENAI_API_KEY)
• --model: LLM model to use (default: gpt-4o)
• --budget: Maximum cost per suggestion in dollars (default: $0.03)
• --prompt, -p: Generate rich context for manual test development
• --debug: Enable detailed debugging output for path resolution
• --json/--yaml: Output results in structured format for automation
• --no-env-check: Skip environment dependency check

Examples:

# Quick heuristic suggestions
aston test-suggest src/calculator.py --k 3

# Rich context for manual test writing
aston test-suggest user/models.py --prompt --yaml context.yaml

# LLM-powered suggestions with cost control
aston test-suggest complex/algorithm.py --llm --budget 0.005

# Target specific function
aston test-suggest "auth/login.py::authenticate_user" --debug

Regression Guard

# Analyze changes for regression risk and get recommendations
aston regression-guard --since HEAD~1 [--until HEAD] [--summary-only]

# Use custom thresholds for development mode
aston regression-guard --since main \
  --max-risk-score 0.8 \
  --max-impacted-nodes 100 \
  --min-test-coverage 0.6 \
  --max-critical-nodes 15

# Generate detailed analysis for CI/CD integration
aston regression-guard --since HEAD~1 \
  --json regression-analysis.json \
  --detailed-output detailed-report.json \
  --exit-code

# Quick summary for development workflow
aston regression-guard --since HEAD~1 --summary-only --quiet

# Disable criticality-based risk assessment for comparison
aston regression-guard --since HEAD~1 --disable-criticality

Intelligent Risk Assessment:

• Multi-Factor Scoring: Combines node count, critical nodes, and test coverage into unified risk score
• Threshold-Based Blocking: Configurable limits prevent high-risk merges automatically
• Test Execution Planning: Prioritized test recommendations based on impact connectivity
• Development vs Production: Flexible thresholds for different workflow contexts
• Criticality-Based Analysis: Enhanced risk assessment based on node importance in the codebase

Regression Detection:

• Change Impact Analysis: Deep call graph traversal to find all affected components
• Critical Node Detection: Identifies high-connectivity components at risk
• Coverage Gap Analysis: Highlights untested code paths in changed areas
• Risk Trend Analysis: Tracks risk patterns across commits and branches

CI/CD Integration:

• Automated Blocking: Exit codes for CI pipeline integration
• Rich Reporting: JSON output for dashboard integration and automation
• Workflow Generation: GitHub Actions and Jenkins pipeline templates
• Configuration Management: YAML/JSON config files for team standards

Options:

• --since: Git reference to analyze changes from (required)
• --until: Git reference to analyze changes to (default: HEAD)
• --max-risk-score: Maximum allowed risk score 0.0-1.0 (default: 0.7)
• --max-impacted-nodes: Maximum allowed impacted nodes (default: 50)
• --min-test-coverage: Minimum required test coverage ratio (default: 0.8)
• --max-critical-nodes: Maximum allowed critical nodes (default: 10)
• --depth: Call graph traversal depth (default: 2)
• --json: Path to write detailed JSON analysis
• --detailed-output: Path to write comprehensive analysis with node details
• --exit-code: Exit with non-zero code if change should be blocked
• --summary-only: Show only summary table, skip detailed output

Risk Thresholds:

• Production Mode: Use default strict thresholds for release branches
• Development Mode: Relax thresholds for active feature development
• Custom Contexts: Team-specific thresholds based on codebase maturity

Examples:

# Standard regression check for PR
aston regression-guard --since main --summary-only

# Development-friendly thresholds
aston regression-guard --since HEAD~1 \
  --max-risk-score 0.9 \
  --max-impacted-nodes 200 \
  --min-test-coverage 0.1

# CI integration with blocking
aston regression-guard --since $BASE_BRANCH \
  --json ci-analysis.json \
  --exit-code

# Detailed analysis for complex changes
aston regression-guard --since feature-branch \
  --detailed-output full-impact.json \
  --depth 3

Environment Check

# Check if all required dependencies are installed
aston check

Options:

• --no-env-check: Skip environment dependency check (also works with any command)

Criticality Analysis

# Analyze code criticality and see top critical components
aston criticality analyze [--top 10] [--min-score 0.5]

# Export criticality scores to JSON for external tools
aston criticality export [--output criticality.json]

# Generate tuned weight configurations based on repository characteristics
aston criticality tune [--output custom_weights.yaml]

Advanced Criticality Metrics:

• Degree Centrality: Measures how connected a component is in the call graph
• Call Depth Analysis: Evaluates the depth of component in the call chain
• Configurable Weights: Customizable importance of different metrics through configuration
• Drift Detection: Identifies changes in component criticality over time
• Impact Assessment: Reveals which components would cause the most damage if broken

Configuration Options:

• --top: Number of critical nodes to display (default: 10)
• --min-score: Minimum criticality score threshold (default: 0.3)
• --output: Path for exporting results in JSON format
• --config: Path to custom weights configuration file
• --detailed: Show additional metrics and explanations

Examples:

# View top 20 most critical components
aston criticality analyze --top 20

# Export criticality data for CI/CD integration
aston criticality export --output criticality.json

# Generate custom weights based on repository structure
aston criticality tune --output custom_weights.yaml

# Use custom configuration with regression guard
aston regression-guard --since main --config custom_weights.yaml

High-Performance Criticality (AstonRank)

# Use the high-performance Go-based criticality scorer (99x faster)
aston criticality analyze --engine astonrank

# Configure algorithms via YAML
aston-rank -config production.yaml -algorithm composite django-graph.json

# Generate sample configuration
aston-rank -generate-config

AstonRank v2.0 Features (Week 3):

• 5 Algorithms: Degree, PageRank, Composite, Betweenness, Eigenvector centrality
• YAML Configuration: Complex algorithm configuration with validation
• Parallel Processing: Multi-threaded algorithms with configurable worker pools
• Sub-Second Performance: 341ms total time for Django dataset (41K nodes, 43K edges)
• Production Ready: Comprehensive error handling, sampling optimization, profiling

Performance Benchmarks:

Algorithm	Time	Memory	Scalability	Parallel
Degree	5ms	32MB	✅ Excellent	✗
PageRank	14ms	64MB	✅ Very Good	✅
Composite	18ms	68MB	✅ Very Good	✅
Betweenness (sampled)	150ms	128MB	⚠️ Good	✅
Eigenvector	25ms	48MB	✅ Very Good	✗

Examples:

# List available algorithms with parameters
aston-rank -list

# Run with configuration file
aston-rank -config configs/production.yaml -algorithm betweenness data.json

# Quick analysis with defaults
aston-rank -algorithm composite -top-k 10 django-graph.json

Local Embedding Generation

# Generate embeddings for all repository files using MiniLM
aston embed --backend minilm

# Generate embeddings for specific files
aston embed --backend minilm --file src/core/*.py

# Preview what would be processed without generating embeddings
aston embed --backend minilm --dry-run

# Force regeneration of existing embeddings
aston embed --backend minilm --force

# Use specific model variant
aston embed --backend minilm --model all-MiniLM-L12-v2

Local Embedding Features:

• MiniLM Integration: Uses sentence-transformers for high-quality local embeddings
• FAISS Vector Store: High-performance similarity search with cosine distance
• Repository-Centric: Stores embeddings in .aston/vectors/<backend>/index.faiss
• Batch Processing: Efficient processing of large codebases with progress tracking
• Filter Integration: Uses unified filter system for intelligent file selection
• No API Dependencies: Completely local processing without external service calls

Options:

• --backend: Embedding provider to use (currently supports: minilm)
• --file: Specific file patterns to process (supports glob patterns)
• --model: Model variant to use (default: all-MiniLM-L6-v2)
• --force: Regenerate embeddings even if they already exist
• --dry-run: Preview file selection without processing

Examples:

# Process entire repository with MiniLM (local)
aston embed --backend minilm

# Process with OpenAI (remote)
aston embed --backend openai

# Auto-select backend (try MiniLM, fallback to OpenAI)
aston embed --backend auto

# Target specific directories
aston embed --backend minilm --file "src/**/*.py" --file "lib/**/*.py"

# OpenAI with custom settings
aston embed --backend openai --model text-embedding-3-large --rate-limit-requests 100

# Quick preview
aston embed --backend minilm --dry-run

# Use larger model for better quality
aston embed --backend minilm --model all-MiniLM-L12-v2

Cache Management

# View cache statistics and performance metrics
aston cache status

# Pre-populate cache for faster operations
aston cache warm-up [--force]

# Clear cache data
aston cache clear [--confirm]

Performance Acceleration:

• Sub-300ms Operations: Ensures all graph operations complete in under 300ms
• Memory Efficient: Optimized 256MB memory limit
• Smart Statistics: Detailed metrics with hit ratios and latency tracking
• Command Integration: Seamless performance enhancement for all analysis commands

Options:

• --force: Force complete cache rebuild during warm-up
• --confirm: Skip confirmation prompt when clearing cache
• --metrics-only: Display only performance metrics in status output
• --verbose: Show detailed cache contents and memory usage

Examples:

# Check cache performance before critical operations
aston cache status --metrics-only

# Force cache rebuild after major changes
aston cache warm-up --force

# Quick cache reset
aston cache clear --confirm

Use Cases

Regression Prevention

Aston's regression guard provides automated protection against code changes that could introduce bugs:

• Pre-Merge Validation: Automatically assess risk before merging pull requests
• Development Workflow: Use relaxed thresholds during active development, strict thresholds for releases
• CI/CD Integration: Block high-risk merges automatically with configurable exit codes
• Test Prioritization: Get prioritized test execution plans based on change impact
• Risk Visualization: Understand change scope with detailed impact analysis and recommendations

Example Workflow:

# Development phase - relaxed thresholds
aston regression-guard --since main \
  --max-risk-score 0.9 \
  --max-impacted-nodes 200 \
  --min-test-coverage 0.1 \
  --summary-only

# Pre-merge validation - strict thresholds (defaults)
aston regression-guard --since main --exit-code

# CI integration with detailed reporting
aston regression-guard --since $BASE_BRANCH \
  --json ci-report.json \
  --exit-code

Test Coverage Analysis

Find and prioritize testing gaps in your codebase:

• Coverage Gap Detection: Identify untested code paths
• Critical Path Analysis: Focus testing on high-impact components
• Test Suggestions: Generate test skeletons with boundary value analysis
• LLM-Powered Suggestions: Advanced test generation for complex scenarios

Code Intelligence

Understand your codebase structure and dependencies:

• Knowledge Graph: Visualize call relationships and imports
• Impact Analysis: Understand change ripple effects
• Repository Analysis: Get comprehensive statistics and insights
• Performance Cache: Sub-300ms graph operations with minimal memory footprint

Repository-Centric Design

Aston follows a repository-centric approach:

• All operations are relative to the repository root (current directory)
• Data is stored in .testindex directory at the repository root
• Path resolution is normalized for consistent matching
• Works with both offline and Neo4j storage

Environment Variables

DEBUG=1                      # Enable debug logging
NEO4J_URI=bolt://localhost:7687  # Optional Neo4j connection (requires astonai[db])
NEO4J_USER=neo4j            # Optional Neo4j username (requires astonai[db])
NEO4J_PASS=password         # Optional Neo4j password (requires astonai[db])
ASTON_NO_ENV_CHECK=1        # Skip environment dependency check
OPENAI_API_KEY=sk-...       # Required for --llm features (requires astonai[llm])

Package Installation Patterns:

• Offline Development: pip install astonai → Use aston init --offline for fast setup
• Live Graph Storage: pip install "astonai[db]" → Use aston init with Neo4j
• Full Features: pip install "astonai[all]" → All capabilities enabled