🚀 Generative AI Mentor Project

This repository documents my journey to mastering Generative AI with a strong focus on .NET applications, OpenAI, and Azure deployment.

---

📌 Features (So Far)

✅ .NET Core Web API backend
✅ Frontend (Bootstrap + Modular JS)
✅ Chat with OpenAI (GPT-3.5, GPT-4o)
✅ Streaming responses with typing animation
✅ Persistent chat history in SQL Server
✅ Multi-session management (create, load, clone, delete)
✅ Prompt templates with parameters (tone, length, style)
✅ Deployed to Azure App Service + Azure SQL
✅ Modularized frontend (chat.js, sessions.js, templates.js, utils.js, main.js)
✅ AdminLTE Panel for managing prompt templates
✅ Versioning & rollback for templates
✅ Cross-model comparison (OpenAI, Gemini, Claude-ready)
✅ Comparison results persisted in SQL Server
✅ AdminLTE pages for running comparisons + viewing history
✅ Side-by-side grid view for provider/model outputs
✅ Deduplication + error persistence (ErrorCode, ErrorMessage)
✅ JWT-secured Admin area for testing
✅ Document ingestion & chunking pipeline (SQL Server)
✅ Embedding generation + vector storage with pgvector (Postgres)
✅ Admin test UI for embeddings + similarity search
✅ Retrieval Service with Semantic, Keyword & Hybrid Search + Admin Debug UI (filter, counters, SLA monitoring, fallback)
✅ RAG pipeline with baseline vs RAG-enhanced answers
✅ DTOs for structured outputs (RagResultDto, RagChunkDto)
✅ Admin Debug UI (RAG Comparison Page with side-by-side answers, context, latency, SLA badge, spinner)
✅ History logging into SQL Server (RagComparisonHistories)
✅ RAG History Page with filters, compare mode, CSV/Excel export
✅ Docs Module for document lifecycle (Upload/Edit/Delete/Rebuild embeddings)
✅ Analytics Dashboard (SLA Compliance, Latencies, PromptStyle Distribution, Trends) with filters + drill-down to history

---

📂 AirNir Project Structure (RAG Implementation)

/AirNir
├── Library
│   ├── ArNir.Core       → Entities, DTOs, Config, Validations
│   ├── ArNir.Data       → DbContexts (SQL Server + Postgres), EF Migrations
│   └── ArNir.Services   → Business logic Service, Interface, Helper, Mapping (EmbeddingService, RetrievalService, RagService)
│
├── Presentation
│   ├── ArNir.Admin      → AdminLTE UI Controllers, ViewModel, Views (embedding + retrieval test pages, RAG comparison, history, analytics)
│   └── ArNir.Frontend   → End-user search/chat interface (future, Phase 3.7+)
│
├── sql
│   ├── create_tables.sql
│   ├── update_documents_chunks.sql
│   ├── update_embeddings.sql
│   └── update_rag_history.sql
│
└── docs
    ├── Phase3_RAG_Architecture.png
    ├── Phase3.3_Architecture.png
    ├── Phase3.4_Architecture.png
    ├── Phase3.5_Architecture.png
    ├── Phase3.6_Architecture.png
    └── GenerativeAI_KnowledgeBase.md

---

📖 Phase 3 Progress

• Phase 3.1 – Document Ingestion & Chunking ✅ Completed
• Phase 3.2 – Embeddings & Vector Storage ✅ Completed
• Phase 3.3 – Retrieval Service ✅ Completed
• Phase 3.4 – Prompt Engineering & RAG Pipeline ✅ Completed
• Phase 3.5 – RAG History Enhancements & Docs Module ✅ Completed
• Phase 3.6 – Analytics & Insights Dashboard ✅ Completed

---

📍 Roadmap

• ✅ Phase 1: Foundation (use cases, HuggingFace models, Azure, prompt engineering)
• ✅ Phase 2: .NET Backend + Frontend Integration (Chat API, sessions, templates, admin panel, comparisons)
• ✅ Phase 3.1: Ingestion Layer (document upload, chunking, SQL Server)
• ✅ Phase 3.2: Embedding & Storage Layer (OpenAI embeddings, Postgres + pgvector)
• ✅ Phase 3.3: Retrieval Service (semantic + hybrid search, Admin debug view)
• ✅ Phase 3.4: Augmentation + Generation Layer (RAG pipeline integration with baseline vs RAG-enhanced answers, Admin Debug UI, history logging)
• ✅ Phase 3.5: Admin Enhancements (History Page, Docs Module)
• ✅ Phase 3.6: Analytics & Insights (Dashboard, Filters, Drill-down)
• ⏳ Phase 3.7: Advanced Analytics (Provider/Model analytics, exports, KPI widgets)

---

🛠️ Tech Stack

• Backend: .NET 9, ASP.NET Core Web API
• Frontend: ASP.NET Core MVC, HTML, Bootstrap, jQuery + Modular JS, AdminLTE
• Database: SQL Server, Postgres (pgvector)
• AI Models: OpenAI GPT-3.5, GPT-4o, Gemini, Claude (planned)
• Cloud: Azure App Service, Azure SQL, Docker (Postgres + pgvector)

---

📌 License

MIT License – feel free to use and adapt!

---

👨‍🏫 Author – pankildesai1988
Learning → Building → Deploying → Scaling 🚀