🧠 Agentic Research AI

An end-to-end Agentic AI system that autonomously plans, executes, evaluates, and completes multi-step research tasks using LLMs (Gemini), FastAPI, and a React + Tailwind frontend, fully deployed on Google Cloud Run.

---

🚀 Live Demo

Cloud Run URL:
👉 https://agentic-ai-539457733412.asia-south1.run.app

---

🧩 What is Agentic Research AI?

Unlike traditional AI chat systems, this project implements a true agentic loop:

1. Planning – Breaks a high-level goal into structured steps
2. Execution – Selects and runs tools autonomously
3. Evaluation – Decides success, retry, or failure
4. State Tracking – Maintains execution state and metrics
5. Termination – Produces a final grounded result

No hard-coded answers. No static flows.

---

🏗️ System Architecture

User Goal
   ↓
Planner (Gemini)
   ↓
Structured Plan (JSON)
   ↓
Executor (Tools)
   ↓
Evaluator (Retry / Fail / Success)
   ↓
State Manager
   ↓
Final Result

Core Tools

• web_search – Research queries
• summarizer – Information synthesis
• python_executor – Controlled computation

---

🖥️ Tech Stack

Backend

• FastAPI
• Google Gemini API
• Python 3.11
• Docker
• Cloud Run

Frontend

• React
• Tailwind CSS
• Framer Motion
• Axios

Infrastructure

• Docker
• Google Cloud Build
• Artifact Registry
• Cloud Run

---

📁 Project Structure

agentic-research-agent/
│
├── backend/
│   ├── api/
│   │   ├── main.py
│   │   ├── agent_runner.py
│   │   └── schemas.py
│   ├── planner/
│   ├── executor/
│   ├── evaluator/
│   ├── state/
│   └── tools/
│
├── frontend/
│   ├── src/
│   ├── public/
│   ├── package.json
|   ├── postcss.config.js
│   └── tailwind.config.js
│
├── Dockerfile
├── .dockerignore
├── .gitignore
└── README.md

---

🧪 Local Development

1️⃣ Backend (FastAPI)

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

pip install -r requirements.txt
uvicorn backend.api.main:app --reload --port 8000

Health check:

http://localhost:8000/api/health

---

2️⃣ Frontend (React)

cd frontend
npm install
npm start

Create production build:

npm run build

---

🐳 Docker (Local)

docker build -t agentic-ai .
docker run -p 8080:8080 -e GEMINI_API_KEY=YOUR_KEY agentic-ai

App runs at:

http://localhost:8080

---

☁️ Deployment (Google Cloud Run)

1️⃣ Create Project & Enable APIs

gcloud projects create agentic-research-ai-prod
gcloud config set project agentic-research-ai-prod
gcloud services enable run.googleapis.com cloudbuild.googleapis.com

---

2️⃣ Build & Push Image

gcloud builds submit --tag asia-south1-docker.pkg.dev/agentic-research-ai-prod/agentic-repo/agentic-ai .

---

3️⃣ Deploy

gcloud run deploy agentic-ai   --image asia-south1-docker.pkg.dev/agentic-research-ai-prod/agentic-repo/agentic-ai   --platform managed   --region asia-south1   --allow-unauthenticated   --set-env-vars GEMINI_API_KEY=YOUR_KEY

---

🔐 Environment Variables

Variable	Description
GEMINI_API_KEY	Google Gemini API key
REACT_APP_BACKEND_URL	Backend base URL (frontend)

---

📈 Metrics & Observability

• Steps executed
• Retry counts
• Tool success rate
• Execution state tracking

---

👨‍💻 Author

Indrasish Bhattacharjee

• GitHub: https://github.com/Indrasish7
• LinkedIn: https://www.linkedin.com/in/indrasishbhattacharjee/

---

📜 License

MIT License

---

⭐ If you like this project, consider giving it a star!