In India there are 3.3 million(approx) registered Non-governmental organizations (NGOs) and many of them face significant operational challenges that hinder their ability to maximize impact with limited resources. These challenges include:

• Repetitive Communication Tasks: Manual handling of donor and volunteer inquiries consuming substantial human resources
• Knowledge Fragmentation: Critical organizational information scattered across multiple sources without centralized access
• Scaling Limitations: Inability to handle increasing volumes of stakeholder interactions without proportional resource increase
• Manual Process Dependencies: Heavy reliance on human intervention for routine communication, data collection, and outreach activities, planning campaigns
• Limited Outreach Capabilities: Difficulty in maintaining consistent and personalized communication across multiple channels
• Data Collection Inefficiencies: Time-intensive manual processes for gathering beneficiary information and feedback

The Sankalpiq Foundation AI Automation Suite is a comprehensive microservices-based Multiagent designed to automate critical NGO operations through intelligent micro agents. The solution addresses operational bottlenecks by implementing specialized micro-agents that handle specific organizational functions while maintaining seamless integration capabilities and working independently.

1. CLI-Based Micro Agent: Intelligent assistant for email automation and knowledge management
2. Voice Micro-Agent: Automated voice interaction system for data collection and FAQ handling
3. WhatsApp Automation Agent: Automated messaging system for scalable outreach campaigns, leads generation
4. Streamlit Dashboard: Web-based user interface for agent overview and solution architecture visualization

https://sankalpiq.streamlit.app

• Response Accuracy: Percentage of queries answered correctly by the knowledge base
• Query Latency: Average response time per query
• Email Delivery Rate: Percentage of successfully delivered emails
• Knowledge Base Coverage: Percentage of responses using stored organizational knowledge

Google Gemini 1.5 Flash serves as the primary language model for this implementation, chosen for the following reasons:

• Cost Efficiency: Optimal balance between performance and operational costs for NGO budgets
• Response Speed: Fast inference times suitable for real-time interactions
• Multilingual Support: Enhanced capability for regional language processing
• Integration Ease: Seamless API integration with existing Google Cloud services
• Context Understanding: Superior performance in understanding organizational context and domain-specific queries

Hugging Face Transformers (all-MiniLM-L6-v2) is utilized for embedding generation and semantic search capabilities:

• Open Source: No licensing costs, suitable for resource-constrained environments
• Offline Capability: Can operate without continuous internet connectivity
• Customization: Ability to fine-tune on organization-specific data
• Privacy: Local processing ensures sensitive organizational data remains secure

• Python 3.8+: Primary development language
• FastAPI: Asynchronous web framework
• Streamlit: Interactive web application framework for dashboard creation
• LangChain: LLM application development framework
• gspread: Google Sheets API integration
• Selenium: Web browser automation
• asyncio: Asynchronous programming support
• python-dotenv: Environment configuration management
• Oauth2client: Handles Google service authentication via Cloud

Each agent operates with its own independent architecture and deployment configuration. For detailed setup instructions, refer to the individual README files in each micro agent's directory.

• Function: Email automation and knowledge base management with semantic search using vector DB and LLM Embeddings
• Interface: Command-line interface for resource-constrained environments
• Setup: See /cli-assistant/README.md for detailed overview,installation and configuration

• Function: Automated voice interactions and data collection via Call
• Sub-Agents: 1. FAQ Agent for query resolution, 2. Info Agent for structured data collection
• Setup: See /voice-micro-agent/README.md for overview, configuration and deployment

• Function: Scalable messaging and outreach automation
• Capabilities: Bulk message sending, delivery tracking, template-based personalization
• Setup: See /whatsapp-micro-agent/README.md for Google Sheets integration, setup, and Overview

• Function: Web-based interface for agent information and solution architecture visualization
• Features: Overall solution overview, architecture diagrams, and agent infrastrucutre
• Setup: See /client/README.md for dashboard configuration and deployment

• CLI Micro Agent Demo: [https://res.cloudinary.com/dxgpsybjw/video/upload/v1748803813/20250602_001104_itnkwd.mp4]
• Voice Agent Interaction Demo: []
• WhatsApp Automation Demo: [https://res.cloudinary.com/dxgpsybjw/video/upload/v1748820409/wp-agent_mw9vut.mp4]
• End-to-End Workflow Demo: [https://sankalpiq.streamlit.app]

• Python 3.8 or higher
• Docker and Docker Compose
• Google Cloud Platform account with API access
• Twilio account with verified phone number
• Gmail account with App Password enabled

# Install dependencies
pip install -r requirements.txt

• Apache Kafka Integration: Implementation of event-driven architecture for real-time data streaming and service decoupling
• Model Context Protocol (MCP) Servers: Deployment on high-performance infrastructure for concurrent task processing
• Cloud-Native Architecture: Migration to containerized deployments on AWS ECS, Google Cloud Run, or Azure Container Instances

• Langflow Integration: Visual workflow orchestration for complex agent behavior design
• Enhanced Language Support: Regional Indian language processing including Bengali, Tamil, and Marathi
• Advanced Analytics: Real-time dashboard for performance monitoring and operational insights

• Multi-Channel Integration: Support for Telegram, and other messaging platforms
• Advanced Speech Recognition: Implementation of specialized STT systems or sarvam-ai for improved regional language accuracy
• Intelligent Routing and Connection: AI-powered query classification and routing to appropriate specialized agents

• Real-Time Alerts: Automated failure detection and notification systems
• Performance Analytics: Comprehensive tracking of agent performance and user engagement metrics
• Audit Logging: Complete interaction logging for compliance and performance optimization

This project maintains high code quality through automated tooling and standardized practices:

• Husky: Pre-commit hooks for code quality enforcement
• Linting: Automated code style checking and formatting
• Version Control: Git-based workflow with branch protection
• Modular Architecture: Each agent maintains independent codebase and deployment configuration
• Documentation Standards: Comprehensive README files for each component
• Detailed Technical Documentation: Detailed technical documentation as per Industry Standards.
• System Design: High Level Architecture and Low Level Architecture made using Mermaid.