🍌 Banana Py Platform - Voice AI Agents Architecture

🟢 Live System Status

🚀 System Running Successfully

Status: ✅ HEALTHY & OPERATIONAL

Server: Running at http://localhost:5050

API Model: gourav-check model via https://01ct5lsg-11434.inc1.devtunnels.ms/api/generate + Google TTS

Platform: Banana Py Platform v2.0.0

Tagline: One Python system satisfy all technical and business requirement

3

Products Available

6

Features Active

100%

System Ready

🏗️ System Architecture Overview

The Banana Py Platform is a comprehensive AI-powered voice system featuring Agent-to-Agent (A2A) communication, Retrieval-Augmented Generation (RAG), multi-AI provider support, and real-time chat capabilities. Built for production deployment on Google Cloud Run with memory optimization and scalable architecture.

✅ Currently Active Features

🤖 A2A RAG System
Agent-to-Agent communication with RAG knowledge base

📚 AR Story Books
Interactive AR-powered storybooks for K8 students

🎮 Learnverse Gaming
Immersive learning environment with gameplay

👩‍🏫 AI Teacher Helper
AI-powered assistant for educators

🎤 Voice Interaction
Real-time voice processing with Google TTS

📊 User Analytics
Comprehensive user interaction tracking

6

AI Providers

4+

Agent Types

50+

API Endpoints

3

Products

1GB

Memory Limit

100%

Cloud Ready

🧩 Core Components

A2A Agent System

Intelligent call routing
Dynamic agent creation
Specialized agents (Customer Service, Sales, Technical Support, Product Expert)
Real-time agent switching
Performance analytics
Memory optimization

Multi-AI Provider System

OpenAI (GPT-4, GPT-3.5)
Anthropic Claude (Sonnet, Haiku, Opus)
Google Gemini (Pro, Flash)
Perplexity AI (Llama-3.1)
Ollama (Local models)
Custom API integration
Automatic fallback
Load balancing

RAG Knowledge Base

Document upload & processing
Intelligent chunking strategies
Semantic search with embeddings
MongoDB storage with fallback
Multi-format support (Text, Code, JSON, CSV)
Context-aware responses
Memory-efficient processing

Voice & Phone System

Twilio integration
Inbound/outbound calls
Google TTS integration
Real-time audio processing
WebSocket media streams
Call routing & management
Voice response optimization

Multi-Client Chat System

Client isolation
Session persistence
Dynamic agent orchestration
Document-aware conversations
Web search integration
Real-time chat rooms
Participant management

LangChain & LangGraph

Advanced RAG workflows
Multi-agent orchestration
Workflow management
Tool integration
State management
Conditional logic
Conversational memory

🤖 AI Providers & Models

OpenAI

GPT-4o, GPT-4o-mini, GPT-4-turbo, GPT-4, GPT-3.5-turbo

Anthropic Claude

Claude-3.5-Sonnet, Claude-3.5-Haiku, Claude-3-Opus, Claude-3-Sonnet

Google Gemini

Gemini-1.5-Pro, Gemini-1.5-Flash, Gemini-1.0-Pro

Perplexity AI

Llama-3.1-Sonar (Small/Large/Huge)

Ollama

Llama3.1:8b, Llama3.1:70b, Mistral:7b, CodeLlama:7b

Custom API

Any custom model integration

🔗 API Endpoints

Core Platform

GET /

Platform information and status

GET /health

System health check with feature status

GET /products

Banana Py products information

POST /chat

Simple chat with intelligent A2A routing

Voice & Phone System

POST /make-call

Initiate outbound calls with A2A processing

POST /make-simple-call

Simple outbound calls without AI processing

GET /call-status/{call_sid}

Get call status and details

A2A Agent System

GET /a2a/status

A2A system status and agent information

GET /a2a/agents

List all available agents

POST /a2a/route-call

Route calls to appropriate agents

GET /a2a/analytics

System analytics and performance metrics

Multi-Client Chat System

POST /multi-client/chat

Multi-client chat with dynamic agent creation

POST /multi-client/documents/upload

Upload documents for RAG processing

POST /multi-client/documents/search

Search uploaded documents

GET /multi-client/documents/{chat_id}

List uploaded documents for a chat session

AI Provider Management

GET /ai-providers/status

Get available AI providers and models

POST /ai-providers/test

Test AI provider connectivity

POST /ai-providers/chat

Chat with specific AI provider and model

LangChain & LangGraph

GET /langchain/status

LangChain system status

POST /langchain/chat

Advanced chat with LangChain workflows

GET /langchain/providers

List available providers for LangChain

Room Management

POST /api/rooms

Create new chat rooms

GET /api/rooms

List all chat rooms

GET /api/rooms/stats

Get room system statistics

API Key Security

GET /api-keys

List all API keys (admin only)

POST /api-keys/generate

Generate new API keys with permissions

POST /api-keys/revoke

Revoke API keys (admin only)

🔄 System Flow & Architecture

Voice Call Processing Flow

1. Call Initiation
User receives call → Twilio webhook → /user-message endpoint

2. A2A Orchestration
Message analyzed → Route to appropriate agent → Dynamic agent creation if needed

3. AI Processing
Agent processes with RAG context → Multi-AI provider selection → Response generation

4. Voice Response
Text-to-speech conversion → Audio streaming → User receives response

Chat & Document Processing Flow

1. Chat Request
User sends message → API authentication → Multi-client routing

2. Document Context
Search uploaded documents → RAG retrieval → Context integration

3. Agent Selection
Analyze requirements → Select/create appropriate agent → Load balancing

4. Response Generation
LangChain/LangGraph processing → Multi-AI provider → Contextual response

🍌 Banana Py Products

AR Story Book

Target: K8 Students
Interactive AR storytelling
Educational content
Age-appropriate themes
Voice-guided learning

Learnverse

Target: Students seeking deep learning
Gamified learning environment
Interactive challenges
Progress tracking
Adaptive difficulty

AI Teacher Helper

Target: Teachers and Educators
Lesson planning assistance
Activity suggestions
Student assessment tools
Curriculum alignment

🚀 Deployment & Configuration

Google Cloud Run Deployment

Optimized for Cloud Run with 1024MiB memory limit and production-ready configuration.

# Quick Deployment
export PROJECT_ID="your-gcp-project-id"
chmod +x deploy_cloudrun.sh
./deploy_cloudrun.sh

# Manual Deployment
gcloud builds submit --tag gcr.io/$PROJECT_ID/banana-py-platform --file Dockerfile .
gcloud run deploy banana-py-platform \
    --image gcr.io/$PROJECT_ID/banana-py-platform \
    --platform managed \
    --region us-central1 \
    --memory 1Gi \
    --cpu 1 \
    --concurrency 10 \
    --max-instances 10 \
    --allow-unauthenticated
                

Environment Variables

# Twilio Configuration
TWILIO_ACCOUNT_SID=your_twilio_account_sid
TWILIO_AUTH_TOKEN=your_twilio_auth_token
TWILIO_PHONE_NUMBER=your_twilio_phone_number

# AI Provider Configuration
OPENAI_API_KEY=sk-your_openai_api_key
ANTHROPIC_API_KEY=sk-ant-your_anthropic_api_key
GEMINI_API_KEY=your_gemini_api_key
PERPLEXITY_API_KEY=pplx-your_perplexity_api_key

# MongoDB (Optional - falls back to in-memory)
MONGODB_URI=mongodb://localhost:27017/

# Deployment
NGROK_URL=your-ngrok-url  # For local development
            

Performance Specifications

1GB

Memory Limit

1

vCPU

10

Concurrency

10

Max Instances

🛡️ Security Features

API Key Authentication

Permission-based access control
Rate limiting (configurable)
Usage tracking and limits
Key expiration dates
Revocation and reactivation

Data Protection

Client isolation
Secure API key handling
Input validation with Pydantic
CORS middleware
Environment variable security

Monitoring & Analytics

Health check endpoints
System status monitoring
Performance analytics
Error tracking
Usage statistics

📚 Documentation & Guides

Setup & Configuration

API Usage

Advanced Features

Testing & Examples

🧪 Live Testing & API Tools

🔗 Quick API Testing

Test your Banana Py Platform endpoints directly from this page:

📊 System Health Check

Endpoint: GET /health

Status: ✅ Available

→ Test Health Endpoint

🍌 Products Information

Endpoint: GET /products

Status: ✅ Available

→ View Products

📖 API Documentation

Endpoint: GET /docs

Status: ✅ Available

→ Open Swagger UI

📋 Alternative Docs

Endpoint: GET /redoc

Status: ✅ Available

→ Open ReDoc

🔑 API Key Authentication

Many endpoints require API key authentication. Here's how to test them:

# Test with API Key (replace 'your_api_key' with actual key)
curl -X GET "http://localhost:5050/a2a-status" \
  -H "X-API-Key: your_api_key"

# Test chat endpoint
curl -X POST "http://localhost:5050/chat" \
  -H "X-API-Key: your_api_key" \
  -H "Content-Type: application/json" \
  -d '{"message": "Hello, test the system"}'

# Test AI providers
curl -X GET "http://localhost:5050/ai-providers/status" \
  -H "X-API-Key: your_api_key"
                

📱 Current System Information

🌐 Server Details

URL: http://localhost:5050

Status: Running

Server: uvicorn

🤖 AI Configuration

Model: gourav-check

API: https://01ct5lsg-11434.inc1.devtunnels.ms/api/generate

TTS: Google Text-to-Speech

📊 System Stats

Products: 3 available

Features: 6 active

Platform: Banana Py v2.0.0

🔍 Real-time Endpoint Testing

Public Endpoints (No Authentication Required)

GET /

✅ TESTED & WORKING - Platform information and status
→ Test Now

GET /health

✅ TESTED & WORKING - System health check with feature status
→ Test Now

GET /products

✅ TESTED & WORKING - Banana Py products information
→ Test Now

GET /docs

✅ AVAILABLE - Interactive API documentation (Swagger UI)
→ Open Swagger UI

GET /redoc

✅ AVAILABLE - Alternative API documentation (ReDoc)
→ Open ReDoc

Protected Endpoints (API Key Required)

Note: These endpoints require API key authentication. Use the format: curl -H "X-API-Key: your_api_key" http://localhost:5050/endpoint

GET /a2a-status

A2A system status and agent information (requires 'a2a' permission)

POST /chat

Simple chat with intelligent A2A routing (requires 'a2a' permission)

GET /ai-providers/status

Get available AI providers and models (requires 'a2a' permission)

POST /make-call

Initiate outbound calls with A2A processing (requires 'calls' permission)