ghosttest 0.2.0

👻 Ghost - AI-Powered Test Generation & Healing for Python

Ghost

Autonomous Test Generation & Self-Healing for Python

Ghost is an intelligent agent that lives in your terminal. It watches your codebase, automatically generates comprehensive test suites, and—most importantly—fixes them when they fail.

---

The Workflow

Ghost operates on a continuous feedback loop, ensuring your tests are always green without manual intervention.

graph TD
    A[File Change Detected] -->|Ghost Watcher| B(Analyze Code)
    B --> C{Tests Exist?}
    C -- No --> D[Generate New Tests]
    C -- Yes --> E[Run Existing Tests]
    D --> E
    E --> F{Tests Pass?}
    F -- Yes --> G[✅ All Green]
    F -- No --> H[Analyze Stack Trace]
    H --> I[Self-Healing Agent]
    I -->|Apply Fix| E

Key Capabilities

Autonomous Generation

Ghost analyzes your Python source code to understand its logic, edge cases, and types. It then generates a complete pytest suite covering happy paths, error conditions, and boundary cases.

Self-Healing Engine

When a test fails, Ghost doesn't just report the error. It captures the stdout, stderr, and stack trace, feeds them back into the LLM, and generates a patch to fix the test code. It repeats this process until the test passes.

Real-time Watcher

Designed to run in the background. Ghost uses watchdog to monitor file system events. As soon as you save a file, Ghost triggers the testing cycle, providing immediate feedback.

Multi-Provider Architecture

Ghost is built to be model-agnostic. It supports:

• Groq: For ultra-low latency feedback loops (recommended for watch mode).
• OpenAI: GPT-4o for complex reasoning and difficult test scenarios.
• Anthropic: Claude 3.5 Sonnet for high-quality, idiomatic code generation.
• Ollama: For local, privacy-focused development using Llama 3 or DeepSeek.

Installation

Using pip (Standard)

pip install ghosttest

Using uv (Fastest)

uv pip install ghosttest

From Source

git clone https://github.com/ghost-team/ghost.git
cd ghost
pip install -e .

Getting Started

1. Initialize Ghost Run the initialization command in your project root. This creates a ghost.toml configuration file.

   ghost init
   

2. Configure API Keys Export your API keys as environment variables or add them to a .env file.

   export OPENAI_API_KEY=sk-...
   # or
   export GROQ_API_KEY=gsk_...
   

3. Start Watching Launch the daemon. Ghost will now monitor your project.

   ghost watch
   

Command Line Interface

Command	Arguments	Description
ghost init	[PATH]	Initializes a new Ghost configuration in the specified directory. Defaults to current directory.
ghost watch	[PATH]	Starts the file watcher daemon. Monitors for changes and triggers the test/heal loop.
ghost generate	<FILE>	Manually triggers test generation for a specific Python file.
ghost config	N/A	Opens an interactive configuration wizard to set providers, models, and paths.
ghost providers	N/A	Lists all currently supported AI providers and available models.
ghost version	N/A	Displays the current installed version of Ghost.

Configuration

Ghost is highly configurable via ghost.toml.

[ghost]
# The AI provider to use (openai, anthropic, groq, ollama)
provider = "openai"

# The specific model identifier
model = "gpt-4o"

# Directory where tests should be generated
test_dir = "tests"

# Maximum number of healing attempts before giving up
max_retries = 3

# List of files or directories to ignore
ignore = [
    "setup.py",
    "migrations/"
]

License

Distributed under the MIT License. See LICENSE for more information.