Unit Test Agent 4j

An AI-powered Java unit test generation agent. Automatically generates high-quality JUnit 5 + Mockito tests.

Features

• 🤖 Multi-LLM Support - OpenAI, Anthropic, Gemini, Zhipu AI
• 🔧 Self-Healing - Auto-fixes compilation errors and test failures
• 📊 Coverage-Driven - Analyzes coverage and supplements missing tests
• 🔄 Iterative Mode - Generates tests one method at a time
• ✅ LSP Syntax Check - Optional semantic analysis before compile
• 🔄 Dynamic Phase Switching - Saves 40-60% tokens (v2.2.0)

Quick Install

Linux / macOS

curl -sSL https://raw.githubusercontent.com/codelogickeep/unit-test-agent-4j/main/build.sh | bash
export PATH="$PATH:$HOME/.utagent"

Windows (PowerShell 7+)

irm https://raw.githubusercontent.com/codelogickeep/unit-test-agent-4j/main/build.ps1 | iex

---

Workflow

┌─────────────┐   ┌─────────────┐   ┌─────────────┐   ┌─────────────┐
│  Analysis   │ → │ Generation  │ → │Verification │ → │   Repair    │
│ Read source │   │ Write tests │   │Compile/Test │   │ Fix errors  │
└─────────────┘   └─────────────┘   └─────────────┘   └─────────────┘
                                            │                │
                                            └────────────────┘
                                              Loop until pass

1. Analysis - Read and analyze source code (AST, dependencies, complexity)
2. Generation - Generate test code based on patterns and knowledge base
3. Verification - Compile, run tests, check coverage
4. Repair - Auto-fix failures, repeat until success or max retries

---

Usage Examples

1. Configure API Key (First Time)

# Interactive config
utagent config --protocol openai --api-key "sk-..." --model "gpt-4o"

# Or use environment variables
export UT_AGENT_API_KEY="sk-..."
export UT_AGENT_MODEL_NAME="gpt-4o"

2. Single File Mode

# Basic
utagent --target src/main/java/com/example/MyService.java

# With knowledge base (learn from existing tests)
utagent --target src/main/java/com/example/MyService.java \
  -kb src/test/java

# Interactive mode (confirm before writes)
utagent --target src/main/java/com/example/MyService.java -i

# Custom coverage threshold
utagent --target src/main/java/com/example/MyService.java --threshold 90

3. Batch Mode (Entire Project)

# Scan and test all uncovered classes
utagent --project /path/to/project

# With exclusions
utagent --project /path/to/project \
  --exclude "**/dto/**,**/vo/**,**/entity/**"

# Dry-run (analyze only)
utagent --project /path/to/project --dry-run

---

Command Line Reference

Main Options

Option	Short	Description	Default
--target <file>	-t	Target Java source file	-
--project <dir>	-p	Project root for batch mode	-
--config <file>	-c	Config file path	auto-detect
--threshold <n>		Coverage threshold (0-100)	80
--interactive	-i	Confirm before file writes	false
--dry-run		Analyze only, no generation	false
--check-env		Check environment and exit	-
-v		Verbose logging	false
-h, --help		Show help	-

LLM Override Options

Option	Description	Example
--protocol	LLM protocol	openai, anthropic, gemini, openai-zhipu
--api-key	API key	sk-...
--base-url	API base URL	https://api.openai.com
--model	Model name	gpt-4o, claude-3-5-sonnet
--temperature	Sampling temperature	0.0 - 1.0
--max-retries	Max retry attempts	5
--save	Save overrides to config	-

Batch Mode Options

Option	Description	Example
--exclude	Exclusion patterns (comma-separated)	**/dto/**,**/vo/**
--dry-run	Analyze only	-

Knowledge Base Options

Option	Short	Description
--knowledge-base	-kb	Path to existing tests for style learning

---

Configuration File (agent.yml)

Create in project root, JAR directory, or ~/.utagent/:

# =============================================================================
# LLM Settings
# =============================================================================
llm:
  protocol: "openai"                      # openai | openai-zhipu | anthropic | gemini
  api-key: "${env:UT_AGENT_API_KEY}"       # Supports environment variables
  base-url: "${env:UT_AGENT_BASE_URL}"     # Optional, protocol default used
  model-name: "gpt-4o"                     # Model name
  temperature: 0.0                        # 0.0 (precise) ~ 1.0 (creative)
  timeout: 120                            # Request timeout (seconds)
  custom-headers: {}                       # Custom HTTP headers

# =============================================================================
# Workflow Settings
# =============================================================================
workflow:
  max-retries: 5                           # Max retry on failure
  coverage-threshold: 80                   # Target coverage (%)
  interactive: false                      # Confirm before writes
  use-lsp: false                          # Enable LSP syntax check
  iterative-mode: false                   # Per-method test generation
  method-coverage-threshold: 80           # Per-method coverage threshold
  skip-low-priority: false                # Skip getters/setters
  max-stale-iterations: 3                 # Stop after N iterations without progress
  min-coverage-gain: 1                    # Min coverage gain (%) per iteration

# =============================================================================
# Batch Mode Settings
# =============================================================================
batch:
  exclude-patterns: ""                     # Glob patterns to exclude
  dry-run: false                           # Analyze only
  
# =============================================================================
# Incremental Mode Settings
# =============================================================================
incremental:
  mode: "uncommitted"                     # uncommitted | staged | compare
  target-ref: "HEAD"                       # Target Git ref

# =============================================================================
# Recommended Dependencies (for environment check)
# =============================================================================
dependencies:
  junit-jupiter: "5.10.1"
  mockito-core: "5.8.0"
  mockito-junit-jupiter: "5.8.0"
  mockito-inline: "5.8.0"
  jacoco-maven-plugin: "0.8.11"

---

Configuration Options Reference

LLM Settings (llm)

Key	Type	Default	Description
protocol	string	"openai"	LLM provider protocol
api-key	string	-	API key (supports ${env:VAR})
base-url	string	-	API base URL
model-name	string	-	Model name
temperature	float	0.0	Sampling temperature
timeout	int	120	Request timeout (seconds)
custom-headers	map	{}	Custom HTTP headers

Workflow Settings (workflow)

Key	Type	Default	Description
max-retries	int	5	Max retry attempts
coverage-threshold	int	80	Target coverage %
interactive	bool	false	Confirm before writes
use-lsp	bool	false	Enable LSP syntax checking
iterative-mode	bool	false	Per-method generation
method-coverage-threshold	int	80	Per-method coverage %
skip-low-priority	bool	false	Skip getters/setters
max-stale-iterations	int	3	Stop after N no-progress iterations
min-coverage-gain	int	1	Min coverage gain % per iteration

Batch Settings (batch)

Key	Type	Default	Description
exclude-patterns	string	""	Comma-separated glob patterns
dry-run	bool	false	Analyze only mode

Incremental Settings (incremental)

Key	Type	Default	Description
mode	string	"uncommitted"	uncommitted / staged / compare
base-ref	string	-	Base Git ref (for compare mode)
target-ref	string	"HEAD"	Target Git ref

---

Supported LLMs

Provider	Protocol	Example Models	Base URL
OpenAI	openai	gpt-4o, gpt-4-turbo	https://api.openai.com
Anthropic	anthropic	claude-3-5-sonnet	https://api.anthropic.com
Google	gemini	gemini-1.5-pro	https://generativelanguage.googleapis.com
Zhipu AI	openai-zhipu	glm-4, glm-4.7	https://open.bigmodel.cn/api/coding/paas/v4
Alibaba	openai	qwen-max	https://dashscope.aliyuncs.com/compatible-mode/v1

---

Environment Variables

Variable	Description
UT_AGENT_API_KEY	LLM API key
UT_AGENT_BASE_URL	LLM API base URL
UT_AGENT_MODEL_NAME	Model name

---

Requirements

• JDK 21+
• Maven 3.8+
• Git

License

Apache License 2.0 - See LICENSE

---

Made with ❤️ for Java developers