memra 0.1.2

A declarative orchestration framework for AI-powered business workflows

Memra SDK

A declarative orchestration framework for AI-powered business workflows. Think of it as "Kubernetes for business logic" where agents are the pods and departments are the deployments.

🚀 Quick Start

from memra import Agent, Department, LLM
from memra.execution import ExecutionEngine

# Define your agents declaratively
data_engineer = Agent(
    role="Data Engineer",
    job="Extract invoice schema from database",
    llm=LLM(model="llama-3.2-11b-vision-preview", temperature=0.1),
    tools=[
        {"name": "DatabaseQueryTool", "hosted_by": "memra"}
    ],
    output_key="invoice_schema"
)

invoice_parser = Agent(
    role="Invoice Parser", 
    job="Extract structured data from invoice PDF",
    llm=LLM(model="llama-3.2-11b-vision-preview", temperature=0.0),
    tools=[
        {"name": "PDFProcessor", "hosted_by": "memra"},
        {"name": "InvoiceExtractionWorkflow", "hosted_by": "memra"}
    ],
    input_keys=["file", "invoice_schema"],
    output_key="invoice_data"
)

# Create a department
ap_department = Department(
    name="Accounts Payable",
    mission="Process invoices accurately into financial system",
    agents=[data_engineer, invoice_parser],
    workflow_order=["Data Engineer", "Invoice Parser"]
)

# Execute the workflow
engine = ExecutionEngine()
result = engine.execute_department(ap_department, {
    "file": "invoice.pdf",
    "connection": "postgresql://user@host/db"
})

if result.success:
    print("✅ Invoice processed successfully!")
    print(f"Data: {result.data}")

📦 Installation

pip install memra-sdk

🔑 Configuration

Step 1: Get Your API Key

Contact [e-mail] to request an API key for early access.

Step 2: Set Your API Key

Once you receive your API key, configure it:

export MEMRA_API_KEY="your-api-key-here"
export MEMRA_API_URL="https://api.memra.co"  # Optional, defaults to production

Or in Python:

import os
os.environ["MEMRA_API_KEY"] = "your-api-key-here"

🎯 Key Features

• Declarative: Define workflows like Kubernetes YAML
• AI-Powered: Built-in LLM integrations for document processing
• Conversational: Agents explain what they're doing in real-time
• Production Ready: Real database connectivity and file processing
• Scalable: Tools execute on Memra's cloud infrastructure

🏗️ Core Concepts

Agents

Agents are the workers in your workflow. Each agent has:

• Role: What they do (e.g., "Data Engineer")
• Job: Their specific responsibility
• Tools: What capabilities they use
• LLM: Their AI model configuration

Departments

Departments coordinate multiple agents:

• Mission: Overall goal
• Workflow Order: Sequence of agent execution
• Manager: Optional oversight and validation

Tools

Tools are hosted capabilities:

• memra: Hosted by Memra (PDF processing, LLMs, databases)
• mcp: Customer-hosted via Model Context Protocol

📊 Example Output

🏢 Starting Accounts Payable Department
📋 Mission: Process invoices accurately into financial system
👥 Team: Data Engineer, Invoice Parser
🔄 Workflow: Data Engineer → Invoice Parser

👤 Data Engineer: Hi! I'm starting my work now...
💭 Data Engineer: My job is to extract invoice schema from database
⚡ Data Engineer: Using tool 1/1: DatabaseQueryTool
✅ Data Engineer: Great! DatabaseQueryTool did real work and gave me useful results
🎉 Data Engineer: Perfect! I completed my work with real data processing

👤 Invoice Parser: Hi! I'm starting my work now...
💭 Invoice Parser: My job is to extract structured data from invoice pdf
⚡ Invoice Parser: Using tool 1/2: PDFProcessor
✅ Invoice Parser: Great! PDFProcessor did real work and gave me useful results

🔍 Tool Discovery

Discover available tools:

from memra import discover_tools, get_api_status

# Check API health
status = get_api_status()
print(f"API Health: {status['api_healthy']}")
print(f"Tools Available: {status['tools_available']}")

# Discover tools
tools = discover_tools()
for tool in tools:
    print(f"- {tool['name']}: {tool['description']}")

📚 Examples

See the examples/ directory for complete workflows:

• accounts_payable_client.py: Invoice processing with database integration
• More examples coming soon!

🆘 Support

• Support: [e-mail]

📄 License

MIT License - see LICENSE file for details.

---

Built with ❤️ by the Memra team Note: The SDK will not work without a valid API key.