# Metis Agent
[](https://badge.fury.io/py/metis-agent)
[](https://www.python.org/downloads/)
[](https://opensource.org/licenses/Apache-2.0)
[](https://pepy.tech/project/metis-agent)
A powerful, modular framework for building AI agents with intelligent memory management and minimal boilerplate code. Metis Agent provides a comprehensive toolkit for creating intelligent agents that can understand user queries, plan and execute complex tasks, and maintain persistent conversations.
**Latest Release: v0.6.0** - Major expansion with secure code execution, advanced tools, and enterprise-grade capabilities.
### What's New in v0.6.0
- **🔒 E2B Code Sandbox** - Secure Python code execution in isolated cloud environments
- **🛠️ 36+ Advanced Tools** - Comprehensive toolkit for development, research, and analysis
- **🔧 Smart Orchestrator** - Intelligent tool selection with parameter extraction
- **📊 Enhanced Analytics** - Advanced memory management with Titans-inspired system
- **🌐 Enterprise Ready** - MCP integration, blueprint system, and production APIs
- **⚡ Performance Optimized** - Improved query analysis and execution strategies
- **🎯 Developer Focused** - Git integration, project management, and automated workflows
## Features
### 🏢 **Core Architecture**
- **Smart Orchestrator**: Intelligent tool selection with parameter extraction and execution strategies
- **Enhanced Memory System**: Titans-inspired adaptive memory with token-aware context management
- **Query Analyzer**: LLM-powered complexity analysis for optimal tool routing
- **Session Management**: Persistent conversations with automatic context preservation
### 🤖 **LLM Integration**
- **Multiple Providers**: OpenAI, Groq, Anthropic, HuggingFace with seamless switching
- **Model Flexibility**: Support for GPT-4, Claude, Llama, Mixtral, and custom models
- **Secure Authentication**: Encrypted API key management with environment fallback
### 🛠️ **Advanced Tool Suite (36+ Tools)**
#### **🔒 Security & Execution**
- **E2B Code Sandbox**: Secure Python execution in isolated cloud environments
- **Bash Tool**: Safe system command execution with output capture
#### **💻 Development Tools**
- **Git Integration**: Complete workflow management (clone, commit, push, merge, etc.)
- **Code Generation**: Multi-language code creation with best practices
- **Unit Test Generator**: Automated test creation with comprehensive coverage
- **Dependency Analyzer**: Project dependency analysis and optimization
- **Project Management**: Full lifecycle management with validation
#### **🔍 Research & Analysis**
- **Deep Research**: Multi-source research with citation management
- **Data Analysis**: Advanced analytics with pandas, numpy, visualization
- **Web Scraper**: Intelligent content extraction with Firecrawl integration
- **Google Search**: Real-time web search with result processing
#### **📝 Content & Communication**
- **Content Generation**: Multi-format content creation (blogs, docs, emails)
- **Text Analyzer**: Advanced NLP analysis with sentiment and entity recognition
- **Blueprint Execution**: Automated workflow and process execution
#### **📁 File & System Operations**
- **File Manager**: Complete file system operations with safety checks
- **Read/Write Tools**: Intelligent file handling with format detection
- **Grep Tool**: Advanced search with regex and pattern matching
### 🌐 **Enterprise Features**
- **MCP Integration**: Model Context Protocol server support
- **Blueprint System**: Workflow automation and process management
- **CLI Interface**: Comprehensive command-line tools for all operations
- **Web API**: RESTful endpoints for integration and automation
- **Memory Analytics**: Real-time performance monitoring and insights
- **Tool Registry**: Dynamic tool discovery and registration system
## Installation
```bash
pip install metis-agent
```
## Starter Templates
Get started quickly with our comprehensive collection of templates and examples:
**[Metis Agent Starter Templates](https://github.com/metis-analytics/metis-starter)** - A complete collection of templates for different use cases:
- **Basic Agent Template** - Simple agent for beginners and quick prototypes
- **Custom Agent Template** - Specialized agents with custom personalities
- **Web App Template** - Flask-based web chat interface
- **Advanced Integration Template** - Enterprise multi-agent systems
- **Custom Tools Template** - Examples for extending agent capabilities
- **Simple Custom Tool Example** - Step-by-step tool development guide
```bash
# Clone the starter templates
git clone https://github.com/metis-analytics/metis-starter.git
cd metis-starter
# Run your first agent
python templates/basic_agent_template.py
```
Each template includes:
- Complete working examples
- Detailed documentation
- Setup instructions
- Customization guides
- Best practices
## Quick Start
### Basic Usage
```python
from metis_agent import SingleAgent
# Create an agent
agent = SingleAgent()
# Process a query
response = agent.process_query("Write a Python function to calculate Fibonacci numbers")
print(response)
```
### Using Different LLM Providers
```python
from metis_agent import SingleAgent, configure_llm
# Configure LLM (OpenAI, Groq, Anthropic, or HuggingFace)
configure_llm("groq", "llama-3.1-8b-instant", "your-api-key")
# Create an agent
agent = SingleAgent()
# Process a query
response = agent.process_query("Explain quantum computing in simple terms")
print(response)
```
### Secure Code Execution with E2B
```python
from metis_agent import SingleAgent
# Create an agent (E2B tool auto-detected)
agent = SingleAgent()
# Execute Python code securely in cloud sandbox
response = agent.process_query("""
Execute this Python code:
```python
import pandas as pd
import matplotlib.pyplot as plt
# Create sample data
data = {'x': [1, 2, 3, 4, 5], 'y': [2, 4, 6, 8, 10]}
df = pd.DataFrame(data)
# Create visualization
plt.figure(figsize=(8, 6))
plt.plot(df['x'], df['y'], marker='o')
plt.title('Sample Data Visualization')
plt.xlabel('X values')
plt.ylabel('Y values')
plt.show()
print(f"Data shape: {df.shape}")
print(df.describe())
```
""")
print(response)
```
### Creating Custom Tools
```python
from metis_agent import SingleAgent, BaseTool, register_tool
class MyCustomTool(BaseTool):
name = "custom_tool"
description = "A custom tool for specialized tasks"
def can_handle(self, task):
return "custom task" in task.lower()
def execute(self, task):
return f"Executed custom tool on: {task}"
# Register the tool
register_tool("custom_tool", MyCustomTool)
# Create an agent
agent = SingleAgent()
# Process a query
response = agent.process_query("Perform a custom task")
print(response)
```
### Using Titans Memory
```python
from metis_agent import SingleAgent
# Create an agent with Titans memory
agent = SingleAgent(use_titans_memory=True)
# Process queries with memory
result1 = agent.process_query("What is machine learning?", session_id="user123")
result2 = agent.process_query("How does it relate to AI?", session_id="user123")
```
## Command Line Interface
Metis Agent provides a comprehensive command-line interface for all operations:
### Core Commands
```bash
# Interactive chat mode
metis chat
# Run a single query
metis run "Write a Python function to calculate Fibonacci numbers"
# Run with specific LLM provider
metis run "Explain quantum computing" --llm groq --model llama-3.1-8b-instant
# Run with memory enabled
metis run "What did we discuss earlier?" --memory --session-id user123
```
### Agent Management
```bash
# Create a new agent configuration
metis agent create --name "CodeExpert" --personality "Expert programming assistant"
# List all configured agents
metis agent list
# Use a specific agent
metis run "Help me debug this code" --agent CodeExpert
# Delete an agent
metis agent delete CodeExpert
```
### API Key Management
```bash
# Set API keys for different providers
metis auth set-key openai sk-your-openai-key
metis auth set-key groq gsk_your-groq-key
metis auth set-key anthropic your-anthropic-key
metis auth set-key e2b your-e2b-api-key
# List configured API keys (shows providers only, not keys)
metis auth list-keys
# Remove an API key
metis auth remove-key openai
# Test API key connectivity
metis auth test openai
```
### E2B Code Sandbox Setup
```bash
# Set E2B API key for secure code execution
metis auth set-key e2b your-e2b-api-key
# Test E2B connectivity
metis auth test e2b
# Execute code in sandbox via CLI
metis chat "Execute this Python code: print('Hello from E2B sandbox!')"
```
### Tool Management
```bash
# List all available tools
metis tools list
# Get detailed information about a tool
metis tools info CodeGenerationTool
# Test a specific tool
metis tools test CodeGenerationTool "Write a hello world function"
# Enable/disable tools
metis tools enable GoogleSearchTool
metis tools disable FirecrawlTool
```
### Memory Operations
```bash
# Show memory statistics
metis memory stats
# Clear memory for a session
metis memory clear --session-id user123
# Export memory to file
metis memory export --output memory_backup.json
# Import memory from file
metis memory import --input memory_backup.json
# Search memory contents
metis memory search "machine learning"
```
### Web Server
```bash
# Start web server with default settings
metis serve
# Start with custom port and memory enabled
metis serve --port 8080 --memory --cors
# Start with specific agent
metis serve --agent CodeExpert --port 5000
# Start with authentication
metis serve --auth --api-key your-server-api-key
```
### Configuration
```bash
# Configure default LLM provider
metis config set-llm --provider groq --model llama-3.1-8b-instant
# Set default memory settings
metis config set-memory --type titans --path ./memory
# View current configuration
metis config show
# Reset configuration to defaults
metis config reset
```
### Development Tools
```bash
# Run system diagnostics
metis dev diagnose
# Test all components
metis dev test
# Generate development templates
metis dev template --type custom-tool --name MyTool
# Profile agent performance
metis dev profile "Complex query for performance testing"
```
## Web Server
Metis Agent includes a web server for API access:
```bash
# Start the web server
metis serve
```
Then make requests to the API:
```bash
curl -X POST http://localhost:5000/api/query \
-H "Content-Type: application/json" \
-d '{"query": "Write a Python function to calculate Fibonacci numbers"}'
```
API Endpoints:
- `GET /` - Server status
- `POST /api/query` - Process a query
- `GET /api/agent-identity` - Get agent information
- `GET /api/memory-insights` - Get memory statistics
- `GET /api/tools` - List available tools
## Detailed Documentation
### Core Components
#### SingleAgent
The main agent class that orchestrates all components:
```python
from metis_agent import SingleAgent
agent = SingleAgent(
use_titans_memory=False, # Enable/disable Titans memory
tools=None, # Custom tools (uses all available if None)
llm_provider="openai", # LLM provider
llm_model=None, # LLM model (uses default if None)
memory_path=None, # Path to memory database
task_file=None # Path to task file
)
```
#### Intent Router
Determines whether a user query is a question or a task:
```python
from metis_agent.core.intent_router import IntentRouter
router = IntentRouter()
intent = router.classify("What is the capital of France?") # Returns "question"
intent = router.classify("Create a Python script to sort a list") # Returns "task"
```
#### Task Manager
Manages tasks and their status:
```python
from metis_agent.core.task_manager import TaskManager
task_manager = TaskManager()
task_manager.add_task("Write a function to calculate Fibonacci numbers")
task_manager.mark_complete("Write a function to calculate Fibonacci numbers")
tasks = task_manager.get_all_tasks()
```
#### Memory Systems
SQLite-based memory:
```python
from metis_agent.memory.sqlite_store import SQLiteMemory
memory = SQLiteMemory("memory.db")
memory.store_input("user123", "What is machine learning?")
memory.store_output("user123", "Machine learning is...")
context = memory.get_context("user123")
```
Titans-inspired adaptive memory:
```python
from metis_agent.memory.titans.titans_memory import TitansInspiredMemory
memory = TitansInspiredMemory("memory_dir")
memory.store_memory("Machine learning is...", "ai_concepts")
relevant_memories = memory.retrieve_relevant_memories("What is deep learning?")
```
### LLM Providers
Configure and use different LLM providers:
```python
from metis_agent.core.llm_interface import configure_llm, get_llm
# Configure LLM
configure_llm("openai", "gpt-4o") # OpenAI
configure_llm("groq", "llama-3.1-8b-instant") # Groq
configure_llm("anthropic", "claude-3-opus-20240229") # Anthropic
configure_llm("huggingface", "mistralai/Mixtral-8x7B-Instruct-v0.1") # HuggingFace
# Get configured LLM
llm = get_llm()
response = llm.chat([{"role": "user", "content": "Hello!"}])
```
### Tools
Available tools (36+ advanced tools):
#### **Security & Execution**
- `E2BCodeSandboxTool`: Secure Python code execution in isolated cloud environments
- `BashTool`: Safe system command execution with output capture
#### **Development & Code**
- `GitIntegrationTool`: Complete Git workflow management
- `CodeGenerationTool`: Multi-language code generation with best practices
- `PythonCodeTool`: Python-specific code analysis and execution
- `UnitTestGeneratorTool`: Automated test creation with comprehensive coverage
- `DependencyAnalyzerTool`: Project dependency analysis and optimization
- `EditTool`: Intelligent code editing with context awareness
#### **Research & Analysis**
- `DeepResearchTool`: Multi-source research with citation management
- `DataAnalysisTool`: Advanced analytics with pandas, numpy, visualization
- `GoogleSearchTool`: Real-time web search with result processing
- `WebScraperTool`: Intelligent content extraction
- `FirecrawlTool`: Advanced web scraping and content analysis
- `TextAnalyzerTool`: NLP analysis with sentiment and entity recognition
#### **Content & Communication**
- `ContentGenerationTool`: Multi-format content creation (blogs, docs, emails)
- `ConversationManagerTool`: Advanced dialogue management
#### **Project & Workflow Management**
- `ProjectManagementTool`: Full project lifecycle management
- `ProjectValidationTool`: Automated project validation and quality checks
- `BlueprintExecutionTool`: Workflow automation and process execution
- `RequirementsAnalysisTool`: Automated requirements gathering and analysis
- `ToolGeneratorTool`: Dynamic tool creation and customization
#### **File & System Operations**
- `FileManagerTool`: Complete file system operations with safety checks
- `FilesystemTool`: Advanced file system navigation and management
- `ReadTool`: Intelligent file reading with format detection
- `WriteTool`: Smart file writing with backup and validation
- `GrepTool`: Advanced search with regex and pattern matching
#### **Mathematical & Scientific**
- `AdvancedMathTool`: Complex mathematical computations and analysis
- `CalculatorTool`: Mathematical calculations with expression parsing
Creating custom tools:
```python
from metis_agent.tools.base import BaseTool
from metis_agent.tools.registry import register_tool
class MyTool(BaseTool):
name = "my_tool"
description = "Custom tool for specific tasks"
def can_handle(self, task):
# Determine if this tool can handle the task
return "specific task" in task.lower()
def execute(self, task):
# Execute the task
return f"Task executed: {task}"
# Register the tool
register_tool("my_tool", MyTool)
```
### API Key Management
Secure storage and retrieval of API keys:
```python
from metis_agent.auth.api_key_manager import APIKeyManager
key_manager = APIKeyManager()
key_manager.set_key("openai", "your-api-key")
api_key = key_manager.get_key("openai")
services = key_manager.list_services()
```
## Testing
Run the comprehensive test suite:
```bash
# Run system tests
python metis_agent/test_system.py
# Run CLI tests
python metis_agent/test_cli.py
```
## Advanced Usage
### Session Management
Maintain context across multiple interactions:
```python
agent = SingleAgent()
# First query
response1 = agent.process_query(
"What are the main types of machine learning?",
session_id="user123"
)
# Follow-up query (uses context from first query)
response2 = agent.process_query(
"Can you explain supervised learning in more detail?",
session_id="user123"
)
```
### Tool Selection
Specify which tool to use for a query:
```python
agent = SingleAgent()
# Use a specific tool
response = agent.process_query(
"Generate a Python function to sort a list",
tool_name="CodeGenerationTool"
)
```
### Memory Insights
Get insights about the agent's memory:
```python
agent = SingleAgent(use_titans_memory=True)
# Process some queries
agent.process_query("What is machine learning?", session_id="user123")
agent.process_query("Explain neural networks", session_id="user123")
# Get memory insights
insights = agent.get_memory_insights()
print(insights)
```
## Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
1. Fork the repository
2. Create your feature branch (`git checkout -b feature/amazing-feature`)
3. Commit your changes (`git commit -m 'Add some amazing feature'`)
4. Push to the branch (`git push origin feature/amazing-feature`)
5. Open a Pull Request
## License
This project is licensed under the Apache License 2.0 - see the [LICENSE](LICENSE) file for details.
## Contact & Links
- **PyPI Package**: [https://pypi.org/project/metis-agent/](https://pypi.org/project/metis-agent/)
- **Starter Templates**: [https://github.com/metis-analytics/metis-starter](https://github.com/metis-analytics/metis-starter)
- **Documentation**: [https://github.com/metis-analytics/metis-agent/wiki](https://github.com/metis-analytics/metis-agent/wiki)
- **Issues & Support**: [https://github.com/metis-analytics/metis-agent/issues](https://github.com/metis-analytics/metis-agent/issues)
- **Discussions**: [https://github.com/metis-analytics/metis-agent/discussions](https://github.com/metis-analytics/metis-agent/discussions)
---
<p align="center">
<strong>Metis Agent - Building Intelligent AI Systems</strong>
</p>
Raw data
{
"_id": null,
"home_page": "https://github.com/metisos/metisos_agentV1",
"name": "metis-agent",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": null,
"keywords": "ai, agent, llm, openai, groq, anthropic, huggingface, ollama, local-models, transformers, tools, memory, conversation, context-management, token-aware, intelligent-summarization, blueprint-detection, autonomous-tools, dynamic-generation, capability-expansion, text-analysis, sentiment-analysis, web-scraping, python-execution, advanced-math, nlp, readability-assessment, e2b, code-sandbox, secure-execution, cloud-sandbox, privacy, offline, quantization, knowledge-base, knowledge-management, sqlite, search, categorization, tagging, import-export, cli-management, knowledge-first-processing",
"author": "Metis OS Team",
"author_email": "cjohnson@metisos.com",
"download_url": "https://files.pythonhosted.org/packages/e4/f0/1453ebaf73c8fc06feb25ce318badd100a5d26a39f08f61476bdfbfe84d7/metis_agent-0.9.5.tar.gz",
"platform": null,
"description": "# Metis Agent\r\n\r\n[](https://badge.fury.io/py/metis-agent)\r\n[](https://www.python.org/downloads/)\r\n[](https://opensource.org/licenses/Apache-2.0)\r\n[](https://pepy.tech/project/metis-agent)\r\n\r\nA powerful, modular framework for building AI agents with intelligent memory management and minimal boilerplate code. Metis Agent provides a comprehensive toolkit for creating intelligent agents that can understand user queries, plan and execute complex tasks, and maintain persistent conversations.\r\n\r\n**Latest Release: v0.6.0** - Major expansion with secure code execution, advanced tools, and enterprise-grade capabilities.\r\n\r\n### What's New in v0.6.0\r\n- **\ud83d\udd12 E2B Code Sandbox** - Secure Python code execution in isolated cloud environments\r\n- **\ud83d\udee0\ufe0f 36+ Advanced Tools** - Comprehensive toolkit for development, research, and analysis\r\n- **\ud83d\udd27 Smart Orchestrator** - Intelligent tool selection with parameter extraction\r\n- **\ud83d\udcca Enhanced Analytics** - Advanced memory management with Titans-inspired system\r\n- **\ud83c\udf10 Enterprise Ready** - MCP integration, blueprint system, and production APIs\r\n- **\u26a1 Performance Optimized** - Improved query analysis and execution strategies\r\n- **\ud83c\udfaf Developer Focused** - Git integration, project management, and automated workflows\r\n\r\n## Features\r\n\r\n### \ud83c\udfe2 **Core Architecture**\r\n- **Smart Orchestrator**: Intelligent tool selection with parameter extraction and execution strategies\r\n- **Enhanced Memory System**: Titans-inspired adaptive memory with token-aware context management\r\n- **Query Analyzer**: LLM-powered complexity analysis for optimal tool routing\r\n- **Session Management**: Persistent conversations with automatic context preservation\r\n\r\n### \ud83e\udd16 **LLM Integration**\r\n- **Multiple Providers**: OpenAI, Groq, Anthropic, HuggingFace with seamless switching\r\n- **Model Flexibility**: Support for GPT-4, Claude, Llama, Mixtral, and custom models\r\n- **Secure Authentication**: Encrypted API key management with environment fallback\r\n\r\n### \ud83d\udee0\ufe0f **Advanced Tool Suite (36+ Tools)**\r\n\r\n#### **\ud83d\udd12 Security & Execution**\r\n- **E2B Code Sandbox**: Secure Python execution in isolated cloud environments\r\n- **Bash Tool**: Safe system command execution with output capture\r\n\r\n#### **\ud83d\udcbb Development Tools**\r\n- **Git Integration**: Complete workflow management (clone, commit, push, merge, etc.)\r\n- **Code Generation**: Multi-language code creation with best practices\r\n- **Unit Test Generator**: Automated test creation with comprehensive coverage\r\n- **Dependency Analyzer**: Project dependency analysis and optimization\r\n- **Project Management**: Full lifecycle management with validation\r\n\r\n#### **\ud83d\udd0d Research & Analysis**\r\n- **Deep Research**: Multi-source research with citation management\r\n- **Data Analysis**: Advanced analytics with pandas, numpy, visualization\r\n- **Web Scraper**: Intelligent content extraction with Firecrawl integration\r\n- **Google Search**: Real-time web search with result processing\r\n\r\n#### **\ud83d\udcdd Content & Communication**\r\n- **Content Generation**: Multi-format content creation (blogs, docs, emails)\r\n- **Text Analyzer**: Advanced NLP analysis with sentiment and entity recognition\r\n- **Blueprint Execution**: Automated workflow and process execution\r\n\r\n#### **\ud83d\udcc1 File & System Operations**\r\n- **File Manager**: Complete file system operations with safety checks\r\n- **Read/Write Tools**: Intelligent file handling with format detection\r\n- **Grep Tool**: Advanced search with regex and pattern matching\r\n\r\n### \ud83c\udf10 **Enterprise Features**\r\n- **MCP Integration**: Model Context Protocol server support\r\n- **Blueprint System**: Workflow automation and process management\r\n- **CLI Interface**: Comprehensive command-line tools for all operations\r\n- **Web API**: RESTful endpoints for integration and automation\r\n- **Memory Analytics**: Real-time performance monitoring and insights\r\n- **Tool Registry**: Dynamic tool discovery and registration system\r\n\r\n## Installation\r\n\r\n```bash\r\npip install metis-agent\r\n```\r\n\r\n## Starter Templates\r\n\r\nGet started quickly with our comprehensive collection of templates and examples:\r\n\r\n**[Metis Agent Starter Templates](https://github.com/metis-analytics/metis-starter)** - A complete collection of templates for different use cases:\r\n\r\n- **Basic Agent Template** - Simple agent for beginners and quick prototypes\r\n- **Custom Agent Template** - Specialized agents with custom personalities\r\n- **Web App Template** - Flask-based web chat interface\r\n- **Advanced Integration Template** - Enterprise multi-agent systems\r\n- **Custom Tools Template** - Examples for extending agent capabilities\r\n- **Simple Custom Tool Example** - Step-by-step tool development guide\r\n\r\n```bash\r\n# Clone the starter templates\r\ngit clone https://github.com/metis-analytics/metis-starter.git\r\ncd metis-starter\r\n\r\n# Run your first agent\r\npython templates/basic_agent_template.py\r\n```\r\n\r\nEach template includes:\r\n- Complete working examples\r\n- Detailed documentation\r\n- Setup instructions\r\n- Customization guides\r\n- Best practices\r\n\r\n## Quick Start\r\n\r\n### Basic Usage\r\n\r\n```python\r\nfrom metis_agent import SingleAgent\r\n\r\n# Create an agent\r\nagent = SingleAgent()\r\n\r\n# Process a query\r\nresponse = agent.process_query(\"Write a Python function to calculate Fibonacci numbers\")\r\nprint(response)\r\n```\r\n\r\n### Using Different LLM Providers\r\n\r\n```python\r\nfrom metis_agent import SingleAgent, configure_llm\r\n\r\n# Configure LLM (OpenAI, Groq, Anthropic, or HuggingFace)\r\nconfigure_llm(\"groq\", \"llama-3.1-8b-instant\", \"your-api-key\")\r\n\r\n# Create an agent\r\nagent = SingleAgent()\r\n\r\n# Process a query\r\nresponse = agent.process_query(\"Explain quantum computing in simple terms\")\r\nprint(response)\r\n```\r\n\r\n### Secure Code Execution with E2B\r\n\r\n```python\r\nfrom metis_agent import SingleAgent\r\n\r\n# Create an agent (E2B tool auto-detected)\r\nagent = SingleAgent()\r\n\r\n# Execute Python code securely in cloud sandbox\r\nresponse = agent.process_query(\"\"\"\r\nExecute this Python code:\r\n```python\r\nimport pandas as pd\r\nimport matplotlib.pyplot as plt\r\n\r\n# Create sample data\r\ndata = {'x': [1, 2, 3, 4, 5], 'y': [2, 4, 6, 8, 10]}\r\ndf = pd.DataFrame(data)\r\n\r\n# Create visualization\r\nplt.figure(figsize=(8, 6))\r\nplt.plot(df['x'], df['y'], marker='o')\r\nplt.title('Sample Data Visualization')\r\nplt.xlabel('X values')\r\nplt.ylabel('Y values')\r\nplt.show()\r\n\r\nprint(f\"Data shape: {df.shape}\")\r\nprint(df.describe())\r\n```\r\n\"\"\")\r\n\r\nprint(response)\r\n```\r\n\r\n### Creating Custom Tools\r\n\r\n```python\r\nfrom metis_agent import SingleAgent, BaseTool, register_tool\r\n\r\nclass MyCustomTool(BaseTool):\r\n name = \"custom_tool\"\r\n description = \"A custom tool for specialized tasks\"\r\n \r\n def can_handle(self, task):\r\n return \"custom task\" in task.lower()\r\n \r\n def execute(self, task):\r\n return f\"Executed custom tool on: {task}\"\r\n\r\n# Register the tool\r\nregister_tool(\"custom_tool\", MyCustomTool)\r\n\r\n# Create an agent\r\nagent = SingleAgent()\r\n\r\n# Process a query\r\nresponse = agent.process_query(\"Perform a custom task\")\r\nprint(response)\r\n```\r\n\r\n### Using Titans Memory\r\n\r\n```python\r\nfrom metis_agent import SingleAgent\r\n\r\n# Create an agent with Titans memory\r\nagent = SingleAgent(use_titans_memory=True)\r\n\r\n# Process queries with memory\r\nresult1 = agent.process_query(\"What is machine learning?\", session_id=\"user123\")\r\nresult2 = agent.process_query(\"How does it relate to AI?\", session_id=\"user123\")\r\n```\r\n\r\n## Command Line Interface\r\n\r\nMetis Agent provides a comprehensive command-line interface for all operations:\r\n\r\n### Core Commands\r\n\r\n```bash\r\n# Interactive chat mode\r\nmetis chat\r\n\r\n# Run a single query\r\nmetis run \"Write a Python function to calculate Fibonacci numbers\"\r\n\r\n# Run with specific LLM provider\r\nmetis run \"Explain quantum computing\" --llm groq --model llama-3.1-8b-instant\r\n\r\n# Run with memory enabled\r\nmetis run \"What did we discuss earlier?\" --memory --session-id user123\r\n```\r\n\r\n### Agent Management\r\n\r\n```bash\r\n# Create a new agent configuration\r\nmetis agent create --name \"CodeExpert\" --personality \"Expert programming assistant\"\r\n\r\n# List all configured agents\r\nmetis agent list\r\n\r\n# Use a specific agent\r\nmetis run \"Help me debug this code\" --agent CodeExpert\r\n\r\n# Delete an agent\r\nmetis agent delete CodeExpert\r\n```\r\n\r\n### API Key Management\r\n\r\n```bash\r\n# Set API keys for different providers\r\nmetis auth set-key openai sk-your-openai-key\r\nmetis auth set-key groq gsk_your-groq-key\r\nmetis auth set-key anthropic your-anthropic-key\r\nmetis auth set-key e2b your-e2b-api-key\r\n\r\n# List configured API keys (shows providers only, not keys)\r\nmetis auth list-keys\r\n\r\n# Remove an API key\r\nmetis auth remove-key openai\r\n\r\n# Test API key connectivity\r\nmetis auth test openai\r\n```\r\n\r\n### E2B Code Sandbox Setup\r\n\r\n```bash\r\n# Set E2B API key for secure code execution\r\nmetis auth set-key e2b your-e2b-api-key\r\n\r\n# Test E2B connectivity\r\nmetis auth test e2b\r\n\r\n# Execute code in sandbox via CLI\r\nmetis chat \"Execute this Python code: print('Hello from E2B sandbox!')\"\r\n```\r\n\r\n### Tool Management\r\n\r\n```bash\r\n# List all available tools\r\nmetis tools list\r\n\r\n# Get detailed information about a tool\r\nmetis tools info CodeGenerationTool\r\n\r\n# Test a specific tool\r\nmetis tools test CodeGenerationTool \"Write a hello world function\"\r\n\r\n# Enable/disable tools\r\nmetis tools enable GoogleSearchTool\r\nmetis tools disable FirecrawlTool\r\n```\r\n\r\n### Memory Operations\r\n\r\n```bash\r\n# Show memory statistics\r\nmetis memory stats\r\n\r\n# Clear memory for a session\r\nmetis memory clear --session-id user123\r\n\r\n# Export memory to file\r\nmetis memory export --output memory_backup.json\r\n\r\n# Import memory from file\r\nmetis memory import --input memory_backup.json\r\n\r\n# Search memory contents\r\nmetis memory search \"machine learning\"\r\n```\r\n\r\n### Web Server\r\n\r\n```bash\r\n# Start web server with default settings\r\nmetis serve\r\n\r\n# Start with custom port and memory enabled\r\nmetis serve --port 8080 --memory --cors\r\n\r\n# Start with specific agent\r\nmetis serve --agent CodeExpert --port 5000\r\n\r\n# Start with authentication\r\nmetis serve --auth --api-key your-server-api-key\r\n```\r\n\r\n### Configuration\r\n\r\n```bash\r\n# Configure default LLM provider\r\nmetis config set-llm --provider groq --model llama-3.1-8b-instant\r\n\r\n# Set default memory settings\r\nmetis config set-memory --type titans --path ./memory\r\n\r\n# View current configuration\r\nmetis config show\r\n\r\n# Reset configuration to defaults\r\nmetis config reset\r\n```\r\n\r\n\r\n### Development Tools\r\n\r\n```bash\r\n# Run system diagnostics\r\nmetis dev diagnose\r\n\r\n# Test all components\r\nmetis dev test\r\n\r\n# Generate development templates\r\nmetis dev template --type custom-tool --name MyTool\r\n\r\n# Profile agent performance\r\nmetis dev profile \"Complex query for performance testing\"\r\n```\r\n\r\n## Web Server\r\n\r\nMetis Agent includes a web server for API access:\r\n\r\n```bash\r\n# Start the web server\r\nmetis serve\r\n```\r\n\r\nThen make requests to the API:\r\n\r\n```bash\r\ncurl -X POST http://localhost:5000/api/query \\\r\n -H \"Content-Type: application/json\" \\\r\n -d '{\"query\": \"Write a Python function to calculate Fibonacci numbers\"}'\r\n```\r\n\r\nAPI Endpoints:\r\n\r\n- `GET /` - Server status\r\n- `POST /api/query` - Process a query\r\n- `GET /api/agent-identity` - Get agent information\r\n- `GET /api/memory-insights` - Get memory statistics\r\n- `GET /api/tools` - List available tools\r\n\r\n## Detailed Documentation\r\n\r\n### Core Components\r\n\r\n#### SingleAgent\r\n\r\nThe main agent class that orchestrates all components:\r\n\r\n```python\r\nfrom metis_agent import SingleAgent\r\n\r\nagent = SingleAgent(\r\n use_titans_memory=False, # Enable/disable Titans memory\r\n tools=None, # Custom tools (uses all available if None)\r\n llm_provider=\"openai\", # LLM provider\r\n llm_model=None, # LLM model (uses default if None)\r\n memory_path=None, # Path to memory database\r\n task_file=None # Path to task file\r\n)\r\n```\r\n\r\n#### Intent Router\r\n\r\nDetermines whether a user query is a question or a task:\r\n\r\n```python\r\nfrom metis_agent.core.intent_router import IntentRouter\r\n\r\nrouter = IntentRouter()\r\nintent = router.classify(\"What is the capital of France?\") # Returns \"question\"\r\nintent = router.classify(\"Create a Python script to sort a list\") # Returns \"task\"\r\n```\r\n\r\n#### Task Manager\r\n\r\nManages tasks and their status:\r\n\r\n```python\r\nfrom metis_agent.core.task_manager import TaskManager\r\n\r\ntask_manager = TaskManager()\r\ntask_manager.add_task(\"Write a function to calculate Fibonacci numbers\")\r\ntask_manager.mark_complete(\"Write a function to calculate Fibonacci numbers\")\r\ntasks = task_manager.get_all_tasks()\r\n```\r\n\r\n#### Memory Systems\r\n\r\nSQLite-based memory:\r\n\r\n```python\r\nfrom metis_agent.memory.sqlite_store import SQLiteMemory\r\n\r\nmemory = SQLiteMemory(\"memory.db\")\r\nmemory.store_input(\"user123\", \"What is machine learning?\")\r\nmemory.store_output(\"user123\", \"Machine learning is...\")\r\ncontext = memory.get_context(\"user123\")\r\n```\r\n\r\nTitans-inspired adaptive memory:\r\n\r\n```python\r\nfrom metis_agent.memory.titans.titans_memory import TitansInspiredMemory\r\n\r\nmemory = TitansInspiredMemory(\"memory_dir\")\r\nmemory.store_memory(\"Machine learning is...\", \"ai_concepts\")\r\nrelevant_memories = memory.retrieve_relevant_memories(\"What is deep learning?\")\r\n```\r\n\r\n### LLM Providers\r\n\r\nConfigure and use different LLM providers:\r\n\r\n```python\r\nfrom metis_agent.core.llm_interface import configure_llm, get_llm\r\n\r\n# Configure LLM\r\nconfigure_llm(\"openai\", \"gpt-4o\") # OpenAI\r\nconfigure_llm(\"groq\", \"llama-3.1-8b-instant\") # Groq\r\nconfigure_llm(\"anthropic\", \"claude-3-opus-20240229\") # Anthropic\r\nconfigure_llm(\"huggingface\", \"mistralai/Mixtral-8x7B-Instruct-v0.1\") # HuggingFace\r\n\r\n# Get configured LLM\r\nllm = get_llm()\r\nresponse = llm.chat([{\"role\": \"user\", \"content\": \"Hello!\"}])\r\n```\r\n\r\n### Tools\r\n\r\nAvailable tools (36+ advanced tools):\r\n\r\n#### **Security & Execution**\r\n- `E2BCodeSandboxTool`: Secure Python code execution in isolated cloud environments\r\n- `BashTool`: Safe system command execution with output capture\r\n\r\n#### **Development & Code**\r\n- `GitIntegrationTool`: Complete Git workflow management\r\n- `CodeGenerationTool`: Multi-language code generation with best practices\r\n- `PythonCodeTool`: Python-specific code analysis and execution\r\n- `UnitTestGeneratorTool`: Automated test creation with comprehensive coverage\r\n- `DependencyAnalyzerTool`: Project dependency analysis and optimization\r\n- `EditTool`: Intelligent code editing with context awareness\r\n\r\n#### **Research & Analysis**\r\n- `DeepResearchTool`: Multi-source research with citation management\r\n- `DataAnalysisTool`: Advanced analytics with pandas, numpy, visualization\r\n- `GoogleSearchTool`: Real-time web search with result processing\r\n- `WebScraperTool`: Intelligent content extraction\r\n- `FirecrawlTool`: Advanced web scraping and content analysis\r\n- `TextAnalyzerTool`: NLP analysis with sentiment and entity recognition\r\n\r\n#### **Content & Communication**\r\n- `ContentGenerationTool`: Multi-format content creation (blogs, docs, emails)\r\n- `ConversationManagerTool`: Advanced dialogue management\r\n\r\n#### **Project & Workflow Management**\r\n- `ProjectManagementTool`: Full project lifecycle management\r\n- `ProjectValidationTool`: Automated project validation and quality checks\r\n- `BlueprintExecutionTool`: Workflow automation and process execution\r\n- `RequirementsAnalysisTool`: Automated requirements gathering and analysis\r\n- `ToolGeneratorTool`: Dynamic tool creation and customization\r\n\r\n#### **File & System Operations**\r\n- `FileManagerTool`: Complete file system operations with safety checks\r\n- `FilesystemTool`: Advanced file system navigation and management\r\n- `ReadTool`: Intelligent file reading with format detection\r\n- `WriteTool`: Smart file writing with backup and validation\r\n- `GrepTool`: Advanced search with regex and pattern matching\r\n\r\n#### **Mathematical & Scientific**\r\n- `AdvancedMathTool`: Complex mathematical computations and analysis\r\n- `CalculatorTool`: Mathematical calculations with expression parsing\r\n\r\nCreating custom tools:\r\n\r\n```python\r\nfrom metis_agent.tools.base import BaseTool\r\nfrom metis_agent.tools.registry import register_tool\r\n\r\nclass MyTool(BaseTool):\r\n name = \"my_tool\"\r\n description = \"Custom tool for specific tasks\"\r\n \r\n def can_handle(self, task):\r\n # Determine if this tool can handle the task\r\n return \"specific task\" in task.lower()\r\n \r\n def execute(self, task):\r\n # Execute the task\r\n return f\"Task executed: {task}\"\r\n\r\n# Register the tool\r\nregister_tool(\"my_tool\", MyTool)\r\n```\r\n\r\n### API Key Management\r\n\r\nSecure storage and retrieval of API keys:\r\n\r\n```python\r\nfrom metis_agent.auth.api_key_manager import APIKeyManager\r\n\r\nkey_manager = APIKeyManager()\r\nkey_manager.set_key(\"openai\", \"your-api-key\")\r\napi_key = key_manager.get_key(\"openai\")\r\nservices = key_manager.list_services()\r\n```\r\n\r\n## Testing\r\n\r\nRun the comprehensive test suite:\r\n\r\n```bash\r\n# Run system tests\r\npython metis_agent/test_system.py\r\n\r\n# Run CLI tests\r\npython metis_agent/test_cli.py\r\n```\r\n\r\n## Advanced Usage\r\n\r\n### Session Management\r\n\r\nMaintain context across multiple interactions:\r\n\r\n```python\r\nagent = SingleAgent()\r\n\r\n# First query\r\nresponse1 = agent.process_query(\r\n \"What are the main types of machine learning?\",\r\n session_id=\"user123\"\r\n)\r\n\r\n# Follow-up query (uses context from first query)\r\nresponse2 = agent.process_query(\r\n \"Can you explain supervised learning in more detail?\",\r\n session_id=\"user123\"\r\n)\r\n```\r\n\r\n### Tool Selection\r\n\r\nSpecify which tool to use for a query:\r\n\r\n```python\r\nagent = SingleAgent()\r\n\r\n# Use a specific tool\r\nresponse = agent.process_query(\r\n \"Generate a Python function to sort a list\",\r\n tool_name=\"CodeGenerationTool\"\r\n)\r\n```\r\n\r\n### Memory Insights\r\n\r\nGet insights about the agent's memory:\r\n\r\n```python\r\nagent = SingleAgent(use_titans_memory=True)\r\n\r\n# Process some queries\r\nagent.process_query(\"What is machine learning?\", session_id=\"user123\")\r\nagent.process_query(\"Explain neural networks\", session_id=\"user123\")\r\n\r\n# Get memory insights\r\ninsights = agent.get_memory_insights()\r\nprint(insights)\r\n```\r\n\r\n## Contributing\r\n\r\nContributions are welcome! Please feel free to submit a Pull Request.\r\n\r\n1. Fork the repository\r\n2. Create your feature branch (`git checkout -b feature/amazing-feature`)\r\n3. Commit your changes (`git commit -m 'Add some amazing feature'`)\r\n4. Push to the branch (`git push origin feature/amazing-feature`)\r\n5. Open a Pull Request\r\n\r\n## License\r\n\r\nThis project is licensed under the Apache License 2.0 - see the [LICENSE](LICENSE) file for details.\r\n\r\n## Contact & Links\r\n\r\n- **PyPI Package**: [https://pypi.org/project/metis-agent/](https://pypi.org/project/metis-agent/)\r\n- **Starter Templates**: [https://github.com/metis-analytics/metis-starter](https://github.com/metis-analytics/metis-starter)\r\n- **Documentation**: [https://github.com/metis-analytics/metis-agent/wiki](https://github.com/metis-analytics/metis-agent/wiki)\r\n- **Issues & Support**: [https://github.com/metis-analytics/metis-agent/issues](https://github.com/metis-analytics/metis-agent/issues)\r\n- **Discussions**: [https://github.com/metis-analytics/metis-agent/discussions](https://github.com/metis-analytics/metis-agent/discussions)\r\n\r\n---\r\n\r\n<p align=\"center\">\r\n <strong>Metis Agent - Building Intelligent AI Systems</strong>\r\n</p>\r\n",
"bugtrack_url": null,
"license": "Apache License 2.0",
"summary": "A comprehensive framework for building powerful AI agents with enhanced tools, Graph-Enhanced Knowledge Base system with relationship intelligence, local model support (HuggingFace/Ollama), autonomous capability expansion, blueprint opportunity detection, and advanced memory management with knowledge-first processing",
"version": "0.9.5",
"project_urls": {
"Bug Tracker": "https://github.com/metisos/metisos_agentV1/issues",
"Documentation": "https://github.com/metisos/metisos_agentV1/blob/main/DOCUMENTATION.md",
"Homepage": "https://github.com/metisos/metisos_agentV1",
"Source Code": "https://github.com/metisos/metisos_agentV1"
},
"split_keywords": [
"ai",
" agent",
" llm",
" openai",
" groq",
" anthropic",
" huggingface",
" ollama",
" local-models",
" transformers",
" tools",
" memory",
" conversation",
" context-management",
" token-aware",
" intelligent-summarization",
" blueprint-detection",
" autonomous-tools",
" dynamic-generation",
" capability-expansion",
" text-analysis",
" sentiment-analysis",
" web-scraping",
" python-execution",
" advanced-math",
" nlp",
" readability-assessment",
" e2b",
" code-sandbox",
" secure-execution",
" cloud-sandbox",
" privacy",
" offline",
" quantization",
" knowledge-base",
" knowledge-management",
" sqlite",
" search",
" categorization",
" tagging",
" import-export",
" cli-management",
" knowledge-first-processing"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "19264c66e3595b9ffe3bfeb51d0d178d52df447dace07a28b2ab4a9b6d8f0143",
"md5": "0fcd55074234594b54ae3ae13cc77723",
"sha256": "2c14e398910ebeb848b07e3634c90771b87eb172026b98e9a0bc3c995a128eca"
},
"downloads": -1,
"filename": "metis_agent-0.9.5-py3-none-any.whl",
"has_sig": false,
"md5_digest": "0fcd55074234594b54ae3ae13cc77723",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.8",
"size": 517115,
"upload_time": "2025-08-09T17:20:38",
"upload_time_iso_8601": "2025-08-09T17:20:38.995455Z",
"url": "https://files.pythonhosted.org/packages/19/26/4c66e3595b9ffe3bfeb51d0d178d52df447dace07a28b2ab4a9b6d8f0143/metis_agent-0.9.5-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "e4f01453ebaf73c8fc06feb25ce318badd100a5d26a39f08f61476bdfbfe84d7",
"md5": "06a72d05489223a027cace9a54f00f90",
"sha256": "795e20d1fd4842b73be65ff32b1590bb62d5a2d9707921dd35d6a5104432a60d"
},
"downloads": -1,
"filename": "metis_agent-0.9.5.tar.gz",
"has_sig": false,
"md5_digest": "06a72d05489223a027cace9a54f00f90",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 454624,
"upload_time": "2025-08-09T17:20:41",
"upload_time_iso_8601": "2025-08-09T17:20:41.464873Z",
"url": "https://files.pythonhosted.org/packages/e4/f0/1453ebaf73c8fc06feb25ce318badd100a5d26a39f08f61476bdfbfe84d7/metis_agent-0.9.5.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-08-09 17:20:41",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "metisos",
"github_project": "metisos_agentV1",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [
{
"name": "metis-agent",
"specs": []
},
{
"name": "flask",
"specs": []
},
{
"name": "flask-cors",
"specs": []
},
{
"name": "requests",
"specs": [
[
">=",
"2.31.0"
]
]
},
{
"name": "pandas",
"specs": [
[
">=",
"2.0.0"
]
]
},
{
"name": "numpy",
"specs": [
[
">=",
"1.24.0"
]
]
},
{
"name": "sqlite3",
"specs": []
},
{
"name": "pytest",
"specs": [
[
">=",
"7.4.0"
]
]
},
{
"name": "pytest-cov",
"specs": [
[
">=",
"4.1.0"
]
]
},
{
"name": "e2b",
"specs": [
[
">=",
"0.1.0"
]
]
},
{
"name": "psutil",
"specs": [
[
">=",
"5.9.0"
]
]
},
{
"name": "python-dotenv",
"specs": [
[
">=",
"1.0.0"
]
]
},
{
"name": "markdown",
"specs": [
[
">=",
"3.4.0"
]
]
}
],
"lcname": "metis-agent"
}
