ai-software-factory 1.0.3

Autonomous AI Software Factory - Production-grade multi-agent system

🤖 AI Software Factory - Claude Code Alternative

A Claude Code-like AI assistant that can read, modify, and generate code in any project. Built with hierarchical multi-agent architecture (Architect → Developer → Auditor) for production-quality software development.

---

✨ Features

🎯 Claude Code-like Interface

• Work in ANY directory - Navigate to any project and start coding
• Full filesystem access - Read, edit, and modify files anywhere
• Directory navigation - /cd, /pwd commands like a shell
• Undo/Redo system - Unlimited undo/redo for all changes
• Automatic backups - Every edit is backed up automatically

🧠 Intelligent Assistance

• Smart intent detection - Automatically distinguishes casual chat from project requests
• Dual mode operation - Switch between chat mode and build mode with /mode
• Clarifying questions - Understands your requirements before building (toggle with /questions)
• Multi-agent workflow - Architect plans, Developer codes, Auditor tests
• Context-aware - Analyzes existing codebase structure and style
• Dynamic Skills Injection - Instantly teach agents new capabilities, rules, and guidelines by adding .md files to the skills/ directory
• Clean terminal output - Human-readable format, no verbose JSON logs

🛡️ Safety First

• Diff previews - Review all changes before applying
• Confirmation prompts - Explicit approval for file modifications
• Backup system - Timestamped backups in .ai-factory-backups/
• Graceful degradation - Works without Docker, pytest, or other tools

💰 Completely Free

• Uses OpenRouter API (free tier available)
• No subscription fees
• No usage limits (depends on your OpenRouter quota)

---

🚀 Quick Start

Prerequisites

• Python 3.13 or higher
• OpenRouter API key (free at openrouter.ai)

Installation

The recommended way to install AI Software Factory securely is via pipx (or pip), which makes the ai-factory command globally available anywhere on your system.

# Install the package globally
pipx install ai-software-factory

# Set your API key as an environment variable
# Mac/Linux:
export OPENROUTER_API_KEY="sk-..."
# Windows PowerShell:
$env:OPENROUTER_API_KEY="sk-..."

Local Development

If you want to contribute to the project or run it locally from the source code:

# Clone the repository
git clone https://github.com/ATANU0023/Ai-factory.git
cd Ai-factory/software-factory

# Install in editable mode
pip install -e .

# Set up environment variables
cp .env.example .env
# Edit .env and add your OPENROUTER_API_KEY

# Launch the factory!
ai-factory

Usage

Option 1: Work on Existing Project (Recommended)

# Navigate to your project
cd C:\Users\YourName\Projects\my-web-app

# Launch AI Factory
ai-factory

💬 You: /pwd
📂 Current directory: C:\Users\YourName\Projects\my-web-app

💬 You: /list
📁 Files (23):
  📄 app/main.py
  📄 app/routes.py
  ...

💬 You: Add error handling to all API endpoints

🤖 Analyzing code...
🤖 Generating modifications...
✅ Modified 3 files

Option 2: Create New Project

ai-factory

💬 You: Create a Flask REST API with user authentication

🤔 Let me ask a few questions...
1. What database should I use? (SQLite/PostgreSQL/MySQL)
   💬 Your answer: SQLite

2. Include JWT authentication?
   💬 Your answer: Yes

✅ Building project...
✅ SUCCESS: Generated 12 files

---

📋 Commands Reference

Command	Description	Example
/help	Show help message	/help
/pwd	Show current directory	/pwd
/cd <path>	Change directory	/cd C:\Projects\app
/list	List files	/list
/read <file>	View file content	/read app.py
/edit <file>	Edit a file	/edit config.py
/undo	Undo last change	/undo
/redo	Redo undone change	/redo
/backups	List backup files	/backups
/questions	Toggle clarifying Q&A	/questions
/mode <type>	Switch mode: chat or build	/mode chat
/status	Show current settings	/status
/ask <prompt>	Generate without questions	/ask Hello world
quit/exit	Exit program	quit

💬 Smart Input Modes

The system intelligently detects whether you're having a conversation or requesting a project:

Casual Conversation (automatically detected):

💬 You: how are u
🤖 I'm here to help build software! What would you like to create?

💬 You: what is today's date?
🤖 Today is Monday, April 06, 2026

💬 You: what can you do?
🤖 I can generate code, modify projects, add features, fix bugs...

Project Requests (automatically detected):

💬 You: create a todo app
🚀 Building your project...

🔄 Mode Switching

Control how the system interprets your input:

# Chat Mode - All input treated as questions
💬 You: /mode chat
✅ Switched to CHAT mode
💬 Chat mode: All inputs treated as questions (use /ask for projects)

💬 You: tell me about Python
🤖 [Conversational response]

💬 You: /ask create a calculator
🚀 Building your project...

# Build Mode - All input treated as project requests (default)
💬 You: /mode build
✅ Switched to BUILD mode

💬 You: create a weather app
🚀 Building your project...

📊 Status Display

Check your current configuration anytime:

💬 You: /status

📊 Current Settings:
   Mode: BUILD
   Clarifying Questions: ON
   Working Directory: C:\Projects\my-app

💡 Use '/mode chat' for conversational mode
💡 Use '/mode build' for project building (default)

---

🏗️ Architecture

Multi-Agent System

User Request
    ↓
┌─────────────┐
│  Architect   │ ← Creates detailed development plan
└──────┬──────┘
       ↓
┌─────────────┐
│  Developer   │ ← Generates code based on plan
└──────┬──────┘
       ↓
┌─────────────┐
│   Auditor    │ ← Tests and validates code
└──────┬──────┘
       ↓
   Success/Fix

Technology Stack

• LangGraph - Agent orchestration and workflow management
• ChromaDB - Vector database for semantic memory and caching
• OpenRouter - Unified LLM API (supports DeepSeek, GPT-4, Claude, etc.)
• Pydantic Settings - Configuration management
• Docker (Optional) - Sandboxed code execution

---

📖 Examples

Example 1: Modify Existing Web App

cd my-flask-project
ai-factory

💬 You: Add rate limiting to all API endpoints

🤖 Reading app/routes.py...
🤖 Analyzing structure...
🤖 Generating modifications...

📋 Preview changes:
--- a/app/routes.py
+++ b/app/routes.py
@@ -5,6 +5,7 @@
 from flask import Flask
+from flask_limiter import Limiter

 @app.route('/api/users')
[e-mail]("100/hour")
 def get_users():
     ...

✅ Apply changes? (yes/no): yes
✅ Modified 1 file
💾 Backup created

Example 2: Add New Feature

💬 You: Add user authentication with JWT

🤖 Analyzing project structure...
🤖 Checking dependencies...
🤖 Generating authentication system...

✅ Created 3 new files:
  • app/auth.py
  • app/models/user.py
  • app/routes/auth_routes.py

✅ Modified 2 files:
  • app/routes.py
  • requirements.txt

Example 3: Fix Bugs

💬 You: Fix the null pointer error in user profile page

🤖 Reading app/routes.py...
🤖 Identifying issue...
🤖 Generating fix...

📋 Fix: Add null check before accessing user.avatar_url
--- a/app/routes.py
+++ b/app/routes.py
@@ -45,7 +45,7 @@
 def profile(user_id):
     user = User.query.get(user_id)
-    avatar = user.avatar_url
+    avatar = user.avatar_url if user else "/static/default.png"
     return render_template('profile.html', avatar=avatar)

✅ Apply fix? (yes/no): yes
✅ Bug fixed!

Example 4: Casual Conversation

# The system automatically detects casual conversation
💬 You: how are u

🤖 I'm here to help build software! Please describe what you'd like to create.

💬 You: what can you do?

🤖 I can generate code, modify projects, add features, fix bugs, write tests, and more!
   Try asking me to build something like "Create a todo app" or "Build a REST API".

# Or switch to chat mode for extended conversations
💬 You: /mode chat
✅ Switched to CHAT mode

💬 You: explain async/await in Python
🤖 [Detailed explanation of async/await...]

💬 You: /mode build
✅ Switched to BUILD mode

---

⚙️ Configuration

Environment Variables

Create a .env file:

# Required
OPENROUTER_API_KEY=sk-or-v1-your-api-key-here

# Optional - Model Configuration
# All models can be customized via environment variables
PLANNING_MODEL_NAME=deepseek/deepseek-chat
PLANNING_MODEL_MAX_TOKENS=8192
PLANNING_MODEL_TEMPERATURE=0.3

CODE_GENERATION_MODEL_NAME=deepseek/deepseek-chat
CODE_GENERATION_MODEL_MAX_TOKENS=8192
CODE_GENERATION_MODEL_TEMPERATURE=0.2

CONVERSATIONAL_MODEL_NAME=deepseek/deepseek-chat
CONVERSATIONAL_MODEL_MAX_TOKENS=500
CONVERSATIONAL_MODEL_TEMPERATURE=0.7

LOG_ANALYSIS_MODEL_NAME=meta-llama/llama-3.1-8b-instruct
LOG_ANALYSIS_MODEL_MAX_TOKENS=2048
LOG_ANALYSIS_MODEL_TEMPERATURE=0.1

SUMMARIZATION_MODEL_NAME=openai/gpt-4o-mini
SUMMARIZATION_MODEL_MAX_TOKENS=2048
SUMMARIZATION_MODEL_TEMPERATURE=0.3

# Other Settings
LOG_LEVEL=INFO
ENABLE_SEMANTIC_CACHE=true
MAX_RETRIES=3

Get your API key at openrouter.ai/keys

Model Selection

Models are configured in config/settings.py and can be overridden via environment variables:

# Default configuration in settings.py
planning_model = ModelConfig(
    model_name="deepseek/deepseek-chat",
    max_tokens=8192,
    temperature=0.3,
)

# Override in .env file
PLANNING_MODEL_NAME=anthropic/claude-3.5-sonnet
PLANNING_MODEL_TEMPERATURE=0.5

See openrouter.ai/models for available models.

Available Models

• Planning Model: Used for architecture design and project planning
• Code Generation Model: Used for writing code
• Conversational Model: Used for Q&A and casual chat
• Log Analysis Model: Used for analyzing execution logs
• Summarization Model: Used for creating summaries

---

🔒 Safety Features

1. Automatic Backups

Every file modification creates a timestamped backup:

.ai-factory-backups/
  ├── app_routes.py.1712345678.bak
  ├── config_py.1712345679.bak
  └── ...

2. Undo/Redo

💬 You: /edit main.py
[Make changes]

💬 You: /undo
✅ Undid changes to main.py

💬 You: /redo
✅ Redid changes to main.py

3. Diff Preview

All edits show unified diff before applying - review exact changes before confirming.

4. Confirmation Prompts

File overwrites require explicit "yes" confirmation. Cancel anytime.

5. Clean Terminal Output

The interactive interface uses human-readable logging:

• Only shows warnings and errors (no verbose JSON logs)
• Clean, minimal output for better user experience
• Detailed JSON logs still available for debugging (non-interactive mode)

---

🆚 Comparison with Other Tools

Feature	AI Factory	Claude Code	Cursor	GitHub Copilot
Cost	FREE	$20/mo	$20/mo	$10/mo
Multi-agent workflow	✅	❌	❌	❌
Smart intent detection	✅	❌	❌	❌
Chat/Build modes	✅	❌	❌	❌
Clarifying questions	✅	❌	❌	❌
Auto backups	✅	❌	❌	❌
Undo/Redo	✅	✅	✅	❌
Clean terminal output	✅	✅	✅	N/A
Open source	✅	❌	❌	❌
Self-hosted	✅	❌	❌	❌

---

🧠 Dynamic Skills System

The AI Software Factory features a dynamic skill system that allows you to instantly customize agent behavior and capabilities without changing any core code.

• How it works: Simply drop a Markdown (.md) file into the skills/ directory.
• Auto-Injection: The SkillManager automatically reads all skill files and injects them as strict instructions into the base prompt of every agent (Architect, Developer, Auditor).
• Use Cases: Perfect for enforcing coding standards, UI guidelines, specific framework patterns, or custom organizational rules.

Example skills/frontend_rules.md:

# Frontend Requirements
- Always use semantic HTML5 tags instead of generic `div`s.
- Keep the user interface clean with plenty of white space.
- Write accessible interfaces (include aria-labels, alt text).

---

🛠️ Development

Project Structure

ai-software-factory/
├── src/ai_software_factory/ # Core Package
│   ├── agents/         # LangGraph AI agents
│   ├── memory/         # ChromaDB & semantic cache
│   ├── sandbox/        # Docker execution environment
│   ├── tools/          # Filesystem/shell capabilities
│   ├── orchestrator/   # LangGraph workflow
│   ├── router/         # Model routing logic
│   ├── observability/  # Logging & telemetry
│   ├── config/         # Configuration
│   ├── ai_factory.py   # Main CLI interface
│   └── main.py         # Core factory logic
├── tests/              # Pytest test suite
├── pyproject.toml      # Package configuration
└── README.md           # This file

Running Tests

# Install test dependencies
pip install pytest pytest-cov

# Run tests
pytest tests/ -v

# With coverage
pytest tests/ --cov=. --cov-report=html

---

📝 License

MIT License - See LICENSE file for details.

---

🙏 Acknowledgments

Inspired by:

• Claude Code by Anthropic
• Cursor IDE
• GitHub Copilot Workspace

Built with:

• LangGraph - Agent orchestration
• ChromaDB - Vector database
• OpenRouter - LLM API gateway

---

📧 Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Email: [e-mail]

---

🌟 Star History

If you find this project useful, please consider giving it a ⭐ star on GitHub!

---

Made with ❤️ by ATANU PRAMANIK