🤖 AgentForge

Your AI team that knows YOUR stuff. Upload documents, connect services, ask questions — AI that actually understands your business.

---

The Problem

Generic AI chatbots don't know your business. They can't read your documents, query your database, or understand your context. Building custom AI agents requires a team of ML engineers and months of work.

The Solution

AgentForge lets anyone create personalized AI agents that:

• Read your documents — Upload PDFs, docs, spreadsheets → AI indexes and understands them
• Remember everything — RAG pipeline with vector search, so AI answers from YOUR data
• Take actions — Not just chat — agents can create tasks, query databases, trigger workflows
• Speak your language — Regional language support for Indian users

---

Architecture

graph TB
    subgraph Frontend
        A[Next.js 15 + React] --> B[Cognito Auth]
        A --> C[Chat Interface]
        A --> D[File Manager]
        A --> E[Agent Builder]
    end
    
    subgraph Backend
        F[FastAPI] --> G[RAG Service]
        F --> H[Chat Service]
        F --> I[Agent Tools]
        G --> J[Qdrant<br/>Vector Search]
        G --> K[AWS Bedrock<br/>Claude Sonnet]
        F --> L[Celery Workers]
        L --> M[File Processing]
    end
    
    subgraph Storage
        N[(PostgreSQL)]
        O[S3<br/>File Storage]
        P[Redis<br/>Task Queue]
    end
    
    A -->|REST API| F
    F --> N
    M --> O
    L --> P

Loading

---

Features

🧠 AI Agents

• Custom Agents — Create agents with specific knowledge bases and personalities
• Pre-built Agents — Ready-to-use agents for common tasks
• Tool System — Extensible tools (summarize, search, analyze, create tasks)
• Agent Registry — Manage and share agents across your team

📄 Document Intelligence (RAG)

• Upload anything — PDF, DOCX, XLSX, CSV, TXT
• Automatic chunking — Smart document splitting for optimal retrieval
• Vector embeddings — 1024-dim embeddings stored in Qdrant
• Semantic search — Ask questions, get answers from your documents
• Folder organization — Organize documents into knowledge bases

💬 Chat

• Multi-session — Multiple chat sessions with different agents
• Context-aware — Chat maintains conversation history
• Tool use — Agents can call tools mid-conversation
• Streaming responses — Real-time token streaming

👥 Multi-tenant

• User authentication — AWS Cognito / NextAuth
• Per-user data isolation — Each user's documents and agents are private
• Onboarding flow — Guided setup for new users
• Pricing tiers — Free → Pro → Business → Enterprise

---

Tech Stack

Layer	Technology	Purpose
Frontend	Next.js 15 + React + TailwindCSS	Modern app with SSR
Backend	FastAPI (Python)	Async API server
LLM	AWS Bedrock (Claude Sonnet)	AI reasoning + generation
Vector DB	Qdrant	Document embeddings + semantic search
Database	PostgreSQL 15	Users, agents, sessions, metadata
Queue	Celery + Redis	Background file processing
Storage	AWS S3	Document file storage
Auth	AWS Cognito / NextAuth	Authentication + authorization
Deployment	Docker Compose	One-command deployment

---

Project Structure

├── frontend/                    # Next.js 15 application
│   ├── src/app/
│   │   ├── (app)/
│   │   │   ├── dashboard/       # Main dashboard
│   │   │   ├── chat/            # Chat interface
│   │   │   ├── agents/          # Agent management
│   │   │   └── files/           # File/folder management
│   │   ├── (auth)/              # Login + signup
│   │   ├── pricing/             # Pricing page
│   │   └── onboarding/          # New user setup
│   ├── src/components/          # Reusable UI components
│   └── src/stores/              # State management
│
├── backend/                     # FastAPI application
│   ├── app/
│   │   ├── api/v1/              # API endpoints (chat, files, agents)
│   │   ├── services/
│   │   │   ├── rag_service.py   # RAG pipeline
│   │   │   ├── chat_service.py  # Chat management
│   │   │   └── bedrock_service.py # LLM integration
│   │   ├── tools/               # Agent tool system
│   │   ├── models/              # SQLAlchemy models
│   │   └── core/                # Config, security
│   └── tests/                   # Unit + integration tests
│
├── docker-compose.yml           # Dev environment
├── docker-compose.prod.yml      # Production setup
├── scripts/                     # Setup + deployment scripts
└── docs/                        # Product spec + architecture

---

Quick Start

Docker (recommended)

git clone https://github.com/abhiFSD/AgentForge.git
cd AgentForge

# Configure
cp backend/.env.example backend/.env
cp frontend/.env.example frontend/.env.local
# Edit with your AWS credentials + Bedrock access

# Launch everything
docker-compose up -d

# Frontend:  http://localhost:3000
# Backend:   http://localhost:8000
# Qdrant:    http://localhost:6333
# PgAdmin:   http://localhost:5050

Manual

# Backend
cd backend
pip install -r requirements.txt
uvicorn app.main:app --reload --port 8000

# Frontend
cd frontend
npm install
npm run dev

---

API Endpoints

Method	Endpoint	Description
POST	/api/v1/chat/sessions	Create chat session
POST	/api/v1/chat/sessions/{id}/messages	Send message (streaming)
GET	/api/v1/agents	List agents
POST	/api/v1/agents	Create custom agent
POST	/api/v1/files/upload	Upload document
GET	/api/v1/files/folders	List folders
POST	/api/v1/rag/query	Semantic search
GET	/health	Health check

---

Pricing Model (Built-in)

Plan	Price	Features
Free	₹0	1 agent, 10 docs, 50 messages/day
Pro	₹499/mo	5 agents, 100 docs, unlimited messages
Business	₹1,499/mo	20 agents, 1000 docs, API access
Enterprise	₹3,999/mo	Unlimited, custom integrations, priority support

---

License

MIT — use it, learn from it, build on it.