Category: Business Operations (Knowledge Base Agent)
-
The Knowledge Base Agent (codenamed Gemini Neural Hub) is a next-generation AI assistant designed to act as a secure, intelligent "Second Brain" for enterprise data. Unlike traditional keyword search tools, it utilizes Retrieval Augmented Generation (RAG) powered by Google's advanced Gemini 1.5/2.5 models to read, understand, and reason across entire PDF documents instantly.
-
This project features a unique "Smart Model Switcher" engine that automatically routes queries to the optimal Google model (Flash, Pro, or Legacy) to ensure 100% uptime and zero 404 errors, wrapped in a stunning Cyberpunk/Sci-Fi interface.
1. The Neural Interface (Dashboard)
- A futuristic HUD for document ingestion and real-time telemetry.
2. Intelligent Reasoning (Chat)
- The agent answering complex queries with structural diagrams.
- 🧠 Smart Brain Engine
-
Auto-Failover: Automatically detects API errors (404/429) and switches between gemini-1.5-flash, gemini-1.5-pro, and gemini-2.0 models instantly.
-
Context-Aware: Injects full document context into the system prompt for high-fidelity answers.
- 🎨 Cyberpunk HUD Interface
-
Immersive Design: Features animated particle backgrounds, glassmorphism panels, and neon typography.
-
Interactive Feedback: Visual and simulated auditory cues (toast notifications) for system status.
- 🖼️ Auto-Diagramming
- Visual Thinking: Intelligently detects requests for "structures", "flows", or "processes" and automatically generates Graphviz/DOT diagrams to visualize the data.
- 📊 Live Telemetry
- Real-Time Stats: A sidebar dashboard showing page counts, character volume, and simulated network latency.
- 🛡️ Enterprise Security
- Ephemeral Processing: API Keys and Document data are processed in-memory and cleared upon session reset. Zero permanent storage.
| Component | Technology | Description |
|---|---|---|
| Frontend | Streamlit | Custom Cyberpunk UI with CSS overrides |
| AI Core | Google Gemini API (Flash, Pro, Legacy) | RAG-style reasoning & diagram generation |
| Backend Logic | Python | Smart model routing, error handling |
| PDF Processing | pypdf |
Fast text extraction |
| Visualization | Graphviz | Auto-rendering workflows & process diagrams |
| Memory Layer | In-Memory | Zero data retention for security |
1) Ingestion Layer: User uploads a PDF. The Streamlit app uses pypdf to extract raw text and metadata.
2) Context Injection: The extracted text is optimized and injected into the System Prompt of the GeminiAgent.
3) Smart Routing Layer: The GeminiBrain attempts connection with the priority model (Flash). If a 404 or connection error occurs, it seamlessly fails over to Pro or Legacy models in real-time.
4) Generation Layer: The active Gemini model generates a markdown response. It decides whether to output text or graphviz code based on the query.
5) Rendering Layer: The Frontend renders the markdown text and compiles any DOT code into visible diagrams instantly.
- Follow these steps to deploy the agent locally:
1) Clone the Repository:
git clone https://github.com/Syed-centem/ai-agent.git
cd ai-agent
2) Install Dependencies: Ensure you have Python installed, then run:
pip install -r requirements.txt
3) Run the Application:
streamlit run app.py
4) Initialize the System:
-> Open the link provided in the terminal (usually http://localhost:8501).
-> Enter your Google Gemini API Key in the secure sidebar HUD.
-> Upload a PDF document.
-> Click the "INITIATE NEURAL LINK" button.
[ ] Voice Integration: Adding Speech-to-Text (STT) for hands-free queries.
[ ] Multi-Modal Analysis: Ability to read charts and images inside PDFs.
[ ] Persistent Memory: Integration with ChromaDB for storing chat history across sessions.