Devendra Ramanuj@devenramanuj153820

### 💡 Project Overview The Hybrid Token-Efficient Routing Agent is an advanced, production-ready backend solution engineered to tackle the challenges of modern LLM deployment: escalating cloud API costs and heavy token usage. Built using FastAPI, this smart routing engine dynamically evaluates incoming user requests and automatically decides the most cost-effective and accurate path for execution. ### ⚙️ How It Works & Architecture Our core routing engine implements a dual-layer intelligence system: 1. **Local Edge Processing (Ollama):** Simple, repetitive, or conversational prompts are automatically routed to a local Ollama instance running `llama3`. This processes requests entirely on local infrastructure, reducing external token dependency to zero. 2. **Cloud Intelligence Layer (Google Gemini 2.5 Flash):** When complex reasoning, multi-turn logic, or coding tasks are detected, the system immediately upgrades the workflow to Google Gemini 2.5 Flash. Gemini's high context window and lightning-fast processing power ensure precise execution for high-priority tasks. ### 🛡️ Failover & Resilience Mechanism One of the key innovations of this project is its robust error handling and failover resiliency. If the local Ollama server drops offline, runs out of memory, or returns an insufficient/short response, the architecture instantly captures the exception. Instead of breaking the user experience, it triggers an automated failover that routes the query straight to Gemini 2.5 Flash, guaranteeing a 100% uptime SLA. ### 🛠️ Technical Implementation * **Backend Framework:** FastAPI (Fully asynchronous implementation) * **Asynchronous Networking:** HTTPX for high-performance non-blocking API calls. * **LLM Models Integration:** Google Gemini 2.5 Flash (via official generative language APIs) and Ollama (`llama3`). * **Environment Architecture:** Configured seamlessly with custom router logic and dynamic environment variable extraction.

To participate in the hackathon, simply click Sign up with AMD. If you are not already a member of the AMD AI Developer Program (ADP), you will need to create an account. ADP members can access AI tutorials, experts, and community support. In addition, new members can claim a free month of DeepLearning.ai pro, $100 in AMD GPU cloud credits, and $50 in Fireworks AI API credits. All hackathon participants will receive on hackathon launch additional compute and API credits for their projects. Ready to build what's next on AMD? Sign up with AMD below to secure your spot.

🌎 Online Hackathon | 📅 August 3–10, 2026| 💻 Powered by native.builder The official native.builder hackathon, hosted by NativelyAI and lablab.ai. Join founders, developers, designers, and builders to create real AI-native applications powered by native.builder: from idea to deployed product, fully in the cloud. 💻 This hackathon is fully online. No travel required.