GenAI-Projects

A collection of production-style Generative AI and AI Agent projects. Each project demonstrates a real-world use case built with modern frameworks — designed to help you understand how to build GenAI products from scratch.

---

At a Glance

Projects	9
LLM Providers	OpenAI GPT · Ollama (llama3.2, local)
Primary Agent Framework	LangGraph (6 of 9 projects)
Languages	Python 3.10+
Deployment	Docker Compose (all backend projects)

---

Projects

Agent Pipelines

Multi-node LangGraph pipelines with FastAPI backends, Streamlit frontends, and SSE streaming.

#	Project	Pipeline	Description
4	Multi-Agent Research System	Source Gatherer → Citation Verifier → Report Writer	Researches a topic from URLs and produces a fully cited Markdown report
6	Regulatory Impact Analysis Agent	Document Parser → Clause Extractor → Industry Classifier → Impact Assessor → Action Plan Generator	Ingests regulatory documents (PDF · URL · text · RSS), extracts clauses, scores severity, and streams compliance action plans
7	Strategic Simulation Agent	Decision Framer → Scenario Generator → Outcome Simulator → Risk Analyzer → Report Generator	Takes any business decision, generates Bull / Base / Bear / Tail Risk scenarios, projects quantitative outcomes, scores risks, and streams a strategic report
8	Knowledge Graph Agent	Document Ingester → Entity Extractor → Relationship Extractor → Graph Builder → Graph Summarizer	Extracts typed entities and relationships from any domain document, builds a queryable NetworkX graph, and enables semantic querying with interactive pyvis visualization
9	Narrative Intelligence Agent	Source Aggregator → Sentiment Analyzer → Narrative Extractor → Pattern Mapper → Intelligence Reporter	Aggregates RSS feeds, URLs, and social text; detects sentiment shifts; classifies narratives as dominant / emerging / contested / fringe; maps meta-patterns (echo chambers, divergence, amplification); streams a strategic intelligence brief

RAG & Retrieval

#	Project	Description
3	Production RAG System	Fully containerized retrieval-augmented generation system with document ingestion, FAISS vector search, chunking strategies, and SSE streaming

Conversational Agents

#	Project	Description
1	Customer Support Agent	Multi-step state machine that categorizes queries, analyzes sentiment, and routes to AI response or human escalation
2	Data Analysis Agent	Conversational agent that translates natural language questions into pandas code and executes them against any CSV

Computer Vision

#	Project	Description
5	Intrusion Detection System	Real-time person detection in user-defined restricted zones with email and desktop alert notifications

---

Full Project Index

#	Project	Stack
1	Customer Support Agent	LangGraph · LangChain · OpenAI
2	Data Analysis Agent	PydanticAI · Pandas · OpenAI
3	Production RAG System	FastAPI · Streamlit · Ollama · FAISS · Docker Compose
4	Multi-Agent Research System	LangGraph · FastAPI · Streamlit · Ollama · Docker Compose
5	Intrusion Detection System	YOLOv11 · OpenCV · SMTP · plyer
6	Regulatory Impact Analysis Agent	LangGraph · FastAPI · Streamlit · Ollama · pdfplumber · feedparser · Docker Compose
7	Strategic Simulation Agent	LangGraph · FastAPI · Streamlit · Ollama · Docker Compose
8	Knowledge Graph Agent	LangGraph · FastAPI · Streamlit · NetworkX · pyvis · Ollama · Docker Compose
9	Narrative Intelligence Agent	LangGraph · FastAPI · Streamlit · Plotly · feedparser · BeautifulSoup4 · Ollama · Docker Compose

---

Getting Started

Each project is self-contained with its own README.md, requirements.txt, and .env.example.

git clone https://github.com/Ramakm/GenAI-Projects.git
cd GenAI-Projects/<project-name>

Prerequisites for LangGraph projects (4, 6, 7, 8, 9)

# Install Ollama and pull the model
curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.2

# Install backend dependencies
cd <project>/backend && pip install -r requirements.txt
uvicorn main:app --reload

# Install frontend dependencies (new terminal)
cd <project>/frontend && pip install -r requirements.txt
streamlit run app.py

Docker (all LangGraph projects)

docker compose up --build -d
docker exec <container>-ollama ollama pull llama3.2

---

Repository Structure

GenAI-Projects/
├── README.md
├── LICENSE
├── Customer-Support-Agent/
├── Data-Analysis-Agent/
├── Production-RAG-System/
├── Multi-Agent-Research-System/
├── Intrusion-Detection-System/
├── Regulatory-Impact-Analysis-Agent/
├── Strategic-Simulation-Agent/
├── Knowledge-Graph-Agent/
└── Narrative-Intelligence-Agent/

---

Tech Stack (across projects)

Category	Technologies
LLM Providers	OpenAI GPT, Ollama (local — llama3.2)
Agent Frameworks	LangGraph, LangChain, PydanticAI
Backend	FastAPI, uvicorn
Frontend	Streamlit
Data Visualization	Plotly
Graph Analysis	NetworkX, pyvis
Vector Search	FAISS
Document & Media Parsing	pdfplumber, feedparser, BeautifulSoup4, httpx
Computer Vision	YOLOv11, OpenCV
Infrastructure	Docker, Docker Compose
Data / Validation	Pydantic, Pandas
Language	Python 3.10+

---

LangGraph Architecture Pattern

Projects 4, 6, 7, 8, and 9 share a consistent architecture:

FastAPI (lifespan + CORS)
  └── LangGraph StateGraph (linear pipeline)
        ├── Node 1..N-1  (sync, LLM JSON calls)
        └── Node N       (streamed token-by-token via Ollama /api/chat)
              ↕ SSE
Streamlit frontend
  ├── N-column agent progress strip
  ├── Streaming report / summary placeholder
  └── Results tabs + downloads

SSE event sequence: agent_start → agent_update (×N) → agent_complete → token (streamed) → done

---

Contributing

Contributions are welcome. To add a new project:

1. Fork the repository
2. Create a new folder with a descriptive name
3. Include a README.md, requirements.txt, and .env.example
4. Submit a pull request

---

License

MIT License. See LICENSE for details.

---

Built by Ramakrushna Mohapatra