3-Tier hybrid AI router that orchestrates FunctionGemma-270M on-device and Gemini 2.5 Flash Lite in the cloud for 99% function-calling accuracy at 548ms avg latency. Built at the Cactus × Google DeepMind Hackathon.

Enables small-to-large self-hosted ai models to use local source code when running tool-calling agentic workloads. We actively data mine 20,900+ (2+ TB) popular github repos using large and small ai models to create reuseable: json, markdown and parquet files for local-first tool-calling models.

Enables small function calling models to find and use 484,000 local source code tools from many open source repos

On-device AI copilot for field technicians — local inference via Cactus + cloud escalation to Gemini.

Surge is a lightweight self-learning AI observability and remediation agent that monitors system and container health and executes validated fixes using historical context and real-time telemetry.

A minimal demo for running Google's Gemma models in Godot 4.5+ via Ollama

Train gemma on your codebase

A comprehensive, production-ready tutorial for fine-tuning Google's FunctionGemma-270M-IT model to build an intelligent E-Commerce Customer Support AI Agent with advanced function calling capabilities.

AI-powered emergency sound detection for people with disabilities. On-device FunctionGemma via Cactus Compute. Sub-50ms response. 82% benchmark score.

A demo app showcasing FunctionGemma 270M running on-device via MLX on Apple Silicon. The app is a simple expense tracker where voice input is parsed into structured function calls by the model — no cloud APIs involved.

A collection of agents that perform different tasks.

🚀 Privacy-first browser AI. Powers the Alibaba Page Agent chatbot using local FunctionGemma models via WebGPU. Fast, secure, and zero-data-leakage on every tab.

A lightweight local sidecar that turns text into structured LIFX command plans