Agentic AI Platform

Agentic AI is an advanced autonomous agent orchestration platform built with Django, Bootstrap 5, and Google Gemini 2.5 Flash API, capable of executing recursive agent pipelines with memory support (RAG) for research, analysis, and automation tasks.

---

🚀 Features

• Multi-agent system for task decomposition, execution, and evaluation.
• RAG-enabled memory with vector search (Qdrant client compatible).
• Gemini 2.5 Flash API integration for content generation and embeddings.
• Interactive Bootstrap 5 front-end with glitchy neon & robotic UI.
• Real-time task status: Planner → Worker → Evaluator → Final Output.
• Dynamic task ID generation and step-wise output visualization.
• Glassmorphic, futuristic UI with animated grid overlay.

---

💻 Tech Stack

• Backend: Django 5.1.5, Python 3.12
• Front-end: Bootstrap 5, HTML/CSS/JS
• AI Models: Google Gemini 2.5 Flash (generateContent, embedContent)
• Vector Database: Qdrant (for memory retrieval)
• Other Libraries: httpx, numpy, pydantic

---

⚡ Installation

1. Clone the repo:

git clone https://github.com/yourusername/agentic-ai.git
cd agentic-ai

2. Create a virtual environment:

python -m venv venv
source venv/bin/activate  # Linux/Mac
venv\Scripts\activate     # Windows

3. Set up environment variables:

Create .env in project root:

GEMINI_API_KEY=your_gemini_2_5_flash_api_key
QDRANT_HOST=localhost
QDRANT_PORT=6333

4. Run Django server:

python manage.py migrate
python manage.py runserver

Open http://127.0.0.1:8000 in your browser.

---

🖥️ Usage

1. Enter a task in the Primary Directive / Task textarea (e.g., "Build a multi-agent protein folding system").
2. Click Execute Agents.
3. Watch the Planner → Worker → Evaluator → Final Output pipeline populate dynamically.

The front-end will display:

• Step-wise agent outputs
• Task ID
• Real-time AI feedback

---

🔗 Gemini 2.5 Flash Integration

• Text generation: generateContent
• Embeddings: embedContent (for RAG / vector memory)
• Make sure to use the correct model name:

EMBED_MODEL = "/gemini-embedding-001"
GEN_MODEL = "models/gemini-2.5-flash"

---

💡 Front-end Highlights

• Neon glitchy headers (Orbitron) and tech-style input forms.
• Glassmorphic cards for task input and results.
• Animated loading spinner and step badges.
• Realistic multi-agent workflow simulation.
• Fully responsive with Bootstrap 5.

---

🛠️ Customization

• Swap Gemini models in agents/client.py for new versions.
• Connect to a live Qdrant instance for RAG memory.
• Extend agent workflows in agents/agent_graph.py.
• Adjust front-end styles in templates/index.html CSS section.

---

⚠️ Notes

• This project is designed for local development; production deployment requires proper security and API key management.
• Make sure Qdrant is running locally or use a cloud instance.
• CSRF protection is disabled for local testing with @csrf_exempt. Re-enable for production.

---

🔮 Future Enhancements

• Streaming agent outputs in real-time.
• Multi-agent collaboration and prioritization.
• Gemini 3+ support for advanced reasoning.
• Interactive dashboard for task analytics and memory inspection.

---