Klarixa ecosistema ia

Inspiration

What it does## Inspiration

Traditional multi-agent frameworks often suffer from high latency, rigid agent routing, and complex deployment pipelines. We were inspired to solve this by creating KLARIXA Ecosystem, an enterprise-grade multi-agent architecture built from the ground up to leverage the multimodal reasoning of Google Gemini alongside scalable, serverless Google Cloud infrastructure. Our goal was to demonstrate how complex, 48-module agent networks can operate efficiently via dynamic, on-demand microservices.

What it does

KLARIXA orchestrates autonomous tasks across an ecosystem of specialized microservices called KWorkers (ranging from OCR processing to WhatsApp gateways and STEM simulation models). At its core is the Nonacortex (5-3-1) orchestration matrix:

• 5 Ingestion Channels: Processes multimodal inputs (text, images, document OCR, telemetry).
• 3 Context & Security Selectors: Evaluates execution risk, validates active subscriptions/tokens, and filters operational intent.
• 1 Master Arbitrator: Powered by Gemini 2.5/3.5, it makes the definitive execution decision and dynamically routes the task to the appropriate KWorker microservice in real time.

How we built it

• Core Engine & SDK: Built using Python 3.11, FastAPI, and the official Google GenAI SDK (google-genai) to integrate Gemini models natively.
• Serverless Infrastructure: Containerized using Docker and deployed on Google Cloud Run for autoscaling execution.
• Multimodal Processing: Integrated Tesseract OCR and Pillow within the backend for real-time document analysis before feeding context to Gemini.
• Monetization & API Management: Connected with Whop webhooks and RevenueCat for enterprise API licensing and in-app subscription management.

Challenges we ran into

1. Latency & Context Routing: Designing a hierarchical 5-3-1 decision loop without adding excessive latency required optimizing system prompts and leveraging fast inference models like Gemini Flash alongside Gemini Pro.
2. Containerized Dependencies: Packaging Tesseract OCR libraries inside lightweight Python Docker containers optimized for Cloud Run required careful build optimization.

Accomplishments that we're proud of

• Successfully implemented the Nonacortex 5-3-1 paradigm using Google's native GenAI SDK.
• Built a zero-downtime, serverless architecture where specialized KWorker tasks execute dynamically on demand.
• Unified enterprise communication (WhatsApp), OCR processing, and advanced reasoning under a single production-ready API on GCP.

What we learned

We gained deep insights into structuring multi-agent systems natively on Google Cloud. We learned how to maximize Gemini's function calling and structured reasoning capabilities to replace fragile rule-based routers with robust, adaptive LLM orchestration.

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Klarixa ecosistema ia

Built With

• docker
• fastapi
• gemini-api
• google-cloud
• google-cloud-run
• google-genai-sdk
• microservices
• python
• rest-api
• revenuecat
• tesseract-ocr
• whatsapp-api