The application implements your entire multi-agent architecture and provides a beautiful two-pane dashboard layout so you can directly interact with it and watch its "Chain of Thought". Key Features Implemented: Multi-Agent Ecosystem: Manager Agent (The Brain): Analyzes the intent of each incoming message and returns a JSON payload directing it to the right department. Sales Agent: Focused on conversions, persuasion, and includes your Auto-Upsell rule to suggest an additional relevant product based on the user's preferences. Operations & Finance Agents: Handle order details and profit tracking queries efficiently. Marketing Agent: Focused on promotional ideas and re-engagement campaigns. Dashboard UI: Left Panel: A WhatsApp-like Chat interface where you can test messages as the customer. It includes a config menu (top right) to dynamically change the Customer Context (Name, Preferences, Past Orders). Right Panel (System Logs): A real-time internal monitoring dashboard. As soon as you hit send, you can see the system retrieve memory, route via the Manager, execute the specific agent, and run the final WhatsApp auto-reply polish script—all step by step. Real-Time Context Injection: Every agent is injected with the smart context via a templating system before execution, giving the LLM deep knowledge of the customer it is responding to. You can try sending queries like: "I want to buy the oversized black tee from your latest collection." (Routes to Sales) "Where is my order from last week?" (Routes to Operations) "What is our revenue for this week?" (Routes to Finance) The app works entirely client-side using the injected Gemini API key, ensuring high responsiveness as you test the logic.
Built With
- 1.3%
- 98.6%
- css
- html
- typescript
Log in or sign up for Devpost to join the conversation.