Sahayak Agent - AI-Powered Teacher Assistant

Overview

Sahayak Agent is an advanced AI-driven multi-agent system designed to comprehensively support teachers with educational tasks. This sophisticated platform leverages Google's Agent Development Kit (ADK) to provide an intelligent teaching assistant that handles everything from curriculum planning and lesson design to student performance analysis and task management.

The system recognizes the multifaceted challenges educators face in managing diverse teaching responsibilities and offers a streamlined, AI-powered approach to enhance teaching effectiveness and productivity.

Key Features

Feature	Description
Interaction Type	Conversational Multi-Agent System
Complexity	Advanced
Agent Type	Hierarchical Multi-Agent Architecture
Components	4 Main Agents + 12 Sub-Agents + Specialized Tools
Vertical	Education
Database	Neo4j Graph Database for Student Analytics
Tools	RAG, Image Generation, Memory Management, Reminders

Agent Architecture

The Sahayak Agent system employs a sophisticated hierarchical multi-agent architecture with the following components:

Root Agent: `TeacherAssistant`

Role: Main orchestrator that coordinates between all sub-agents
Model: gemini-2.5-flash
Description: Primary interface for teachers, routes requests to appropriate specialized agents

Main Agents

1. Planner Agent (`planner_agent`)

Purpose: Comprehensive educational planning and content creation

Sub-Agents:
- curriculum_planner: Creates yearly academic plans with calendar integration
- lesson_designer: Develops detailed lesson plans and teaching aids
- content_creator: Orchestrates educational material production
- presentation_generator: Creates interactive slides with visual aids
- questions_generator: Generates quizzes and assessments
- interactive_whiteboard: Creates diagrams and flowcharts

Key Capabilities:

Yearly curriculum planning with local context integration
Detailed lesson plan creation with differentiation support
Interactive presentation slide generation
Assessment and quiz creation
Visual content generation (diagrams, flowcharts)

2. Query Agent (`query_agent`)

Purpose: Content queries and document-based inquiries

Sub-Agents:
- textbook_content_agent: RAG-based textbook queries with analogies
- interactive_image_agent: Image fetching and marked section analysis
- flow_diagram_agent: Process visualization and concept mapping

Key Capabilities:

Textbook content retrieval using RAG (Retrieval-Augmented Generation)
Interactive image analysis with marked section explanations
Flow diagram and process chart creation
Concept explanation with relatable analogies

3. Academia Agent (`academia_agent`)

Purpose: Student performance analysis and academic insights

Sub-Agents:
- student_performance_analyzer: Neo4j database queries for academic analytics

Key Capabilities:

Student performance tracking and analysis
Study team formation based on complementary strengths
Academic statistics and trend generation
Data-driven educational recommendations

4. Task Agent (`task_agent`)

Purpose: Schedule management and reminder system Key Capabilities:

Timetable management and schedule queries
Smart reminder creation with context awareness
Task organization and prioritization
Friendly conversational support for daily planning

Tools Integration

Core Tools

RAG Tools: rag_query, ask_vertex_retrieval for content retrieval
Image Generation: create_slide_images, fetch_textbook_image for visual content
Memory Management: memorize, memorize_dict, memorize_list for state persistence
Graph Analytics: GraphVisualizer for Neo4j student performance queries
Reminder System: create_reminder for intelligent scheduling

Database Integration

Neo4j Graph Database: Stores student performance data, relationships, and academic metrics
Vertex AI: Powers the RAG system for textbook content retrieval
Google Cloud Storage: Handles artifact and image storage

Setup and Installation

Prerequisites

Python 3.11+
Node.js (for Neo4j dependencies)
Google Cloud Platform project
Google Cloud CLI

Installation

Clone the repository:

git clone https://github.com/overfitagents/sahayak-agent.git
cd sahayak-agent

Set up Python environment:

python -m venv env
.\env\Scripts\activate
pip install -r requirements.txt

Install Node.js dependencies:

npm install

Configuration

Set up Google Cloud credentials:

$env:GOOGLE_GENAI_USE_VERTEXAI="true"
$env:GOOGLE_CLOUD_PROJECT="<your-project-id>"
$env:GOOGLE_CLOUD_LOCATION="<your-project-location>"
$env:GOOGLE_CLOUD_STORAGE_BUCKET="<your-storage-bucket>"

Authenticate your Google Cloud account:

gcloud auth application-default login
gcloud auth application-default set-quota-project $env:GOOGLE_CLOUD_PROJECT

Configure Neo4j (for student performance analytics):

# Set your Neo4j connection details in environment variables
$env:NEO4J_URI="bolt://localhost:7687"
$env:NEO4J_USER="neo4j"
$env:NEO4J_PASSWORD="your-password"

Running the Agent

Using Google ADK

The Agent Development Kit provides convenient ways to interact with Sahayak Agent:

CLI Interface

adk run sahayak.agent:root_agent

Web Interface

adk web

The web command starts a local server. Open the provided URL, select "TeacherAssistant" from the dropdown, and interact through the chatbot interface.

FastAPI Server

python main.py

This starts the FastAPI server on http://localhost:8080 with a web interface.

Example Interactions

Curriculum Planning

Teacher: "Create a yearly curriculum plan for Grade 8 Mathematics"
→ Routes to planner_agent → curriculum_planner
→ Analyzes grade level, calendar, and local context
→ Generates comprehensive yearly plan with monthly breakdowns
→ Includes learning objectives, activities, and assessments

Lesson Design

Teacher: "Design a lesson plan for photosynthesis chapter"
→ Routes to planner_agent → lesson_designer
→ Uses RAG to retrieve photosynthesis content
→ Creates detailed lesson with activities and differentiation
→ Generates JSON-structured lesson plan

Content Creation

Teacher: "Create presentation slides for cell division"
→ Routes to planner_agent → content_creator → presentation_generator
→ Generates slide contents based on lesson plan
→ Creates visual images for each slide
→ Provides engaging, localized content

Student Performance Analysis

Teacher: "Form study teams for Grade 6 students in Science topics"
→ Routes to academia_agent → student_performance_analyzer
→ Queries Neo4j database for student performance
→ Creates balanced teams based on strengths/weaknesses
→ Returns structured JSON with team recommendations

Content Queries

Teacher: "Explain photosynthesis with a simple analogy"
→ Routes to query_agent → textbook_content_agent
→ Uses RAG to retrieve content
→ Generates kitchen/factory analogies
→ Provides comprehensive explanation with examples

Task Management

Teacher: "Remind me to prepare tomorrow's class materials"
→ Routes to task_agent
→ Creates intelligent reminder based on schedule
→ Sets appropriate timing (5-6 hours before deadline)
→ Maintains context for follow-up conversations

Usage Workflows

1. Curriculum to Content Pipeline

Curriculum Planning → Lesson Design → Content Creation → Assessment Generation

2. Query and Analysis Flow

Content Query → RAG Retrieval → Analysis → Explanation with Analogies

3. Student Analytics Workflow

Performance Query → Neo4j Analysis → Team Formation → Educational Recommendations

Development and Testing

Running Tests

# Activate virtual environment
.\env\Scripts\activate

# Run basic functionality tests
python -m pytest tests/ -v

# Test agent evaluation
python -m pytest eval/ -v

Development Mode

# Start with hot reload for development
uvicorn main:app --reload --host 127.0.0.1 --port 8080

Deployment

Google Cloud Deployment

# Build and deploy to Google Cloud
gcloud app deploy app.yaml

# Or use Cloud Run
gcloud run deploy sahayak-agent --source .

Docker Deployment

# Build Docker image
docker build -t sahayak-agent .

# Run container
docker run -p 8080:8080 sahayak-agent

Customization

Adding New Sub-Agents

Create agent in appropriate subagent directory
Define tools and instructions
Register with parent agent
Update routing logic in coordinator

Extending RAG Capabilities

Add new content sources to Vertex AI
Update query tools with new retrieval patterns
Enhance content processing pipelines

Custom Analytics

Extend Neo4j schema for new data types
Add query patterns in GraphVisualizer tool
Create new analysis agents for specific metrics

Project Structure

sahayak-agent/
├── main.py                     # FastAPI application entry point
├── requirements.txt            # Python dependencies
├── package.json               # Node.js dependencies
├── Dockerfile                 # Container configuration
├── sahayak/
│   ├── agent.py               # Root agent definition
│   ├── prompt.py              # Core instruction prompts
│   ├── subagents/
│   │   ├── planner/           # Planning and content creation agents
│   │   ├── query/             # Content query and retrieval agents
│   │   ├── academia/          # Student performance analysis agents
│   │   └── reminder/          # Task and reminder management
│   ├── tools/
│   │   ├── rag.py            # RAG retrieval tools
│   │   ├── image.py          # Image generation tools
│   │   ├── graph.py          # Neo4j database tools
│   │   ├── memory.py         # State management tools
│   │   └── reminder.py       # Reminder system tools
│   └── shared_libs/
│       └── types.py          # Shared data types and schemas
├── generated_images/          # Auto-generated visual content
├── interactive_images/        # Interactive teaching materials
└── env/                      # Python virtual environment

Contributing

Fork the repository
Create a feature branch
Implement changes with appropriate tests
Submit a pull request with detailed description

License

This project is licensed under the Apache License 2.0. See LICENSE file for details.

Support

For support and questions:

Create an issue in the GitHub repository
Review the documentation in each subagent's README
Check the example interactions for usage patterns

Sahayak Agent - Empowering educators with intelligent, multi-faceted AI assistance for comprehensive teaching support.

Name		Name	Last commit message	Last commit date
sahayak		sahayak
.gitignore		.gitignore
Dockerfile		Dockerfile
README.md		README.md
cloudbuild.yaml		cloudbuild.yaml
initial_state.json		initial_state.json
main.py		main.py
multiagent.png		multiagent.png
package-lock.json		package-lock.json
package.json		package.json
requirements.txt		requirements.txt
student_graph.py		student_graph.py

Uh oh!

Latest commit

History

Folders and files

Repository files navigation

Sahayak Agent - AI-Powered Teacher Assistant

Overview

Key Features

Agent Architecture

Root Agent: TeacherAssistant

Main Agents

1. Planner Agent (planner_agent)

2. Query Agent (query_agent)

3. Academia Agent (academia_agent)

4. Task Agent (task_agent)

Tools Integration

Core Tools

Database Integration

Setup and Installation

Prerequisites

Installation

Configuration

Running the Agent

Using Google ADK

CLI Interface

Web Interface

FastAPI Server

Example Interactions

Curriculum Planning

Lesson Design

Content Creation

Student Performance Analysis

Content Queries

Task Management

Usage Workflows

1. Curriculum to Content Pipeline

2. Query and Analysis Flow

3. Student Analytics Workflow

Development and Testing

Running Tests

Development Mode

Deployment

Google Cloud Deployment

Docker Deployment

Customization

Adding New Sub-Agents

Extending RAG Capabilities

Custom Analytics

Project Structure

Contributing

License

Support

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

Root Agent: `TeacherAssistant`

1. Planner Agent (`planner_agent`)

2. Query Agent (`query_agent`)

3. Academia Agent (`academia_agent`)

4. Task Agent (`task_agent`)