Marco Somma

What OrKa does (Maker’s description): OrKa is a no-BS, YAML-first orchestrator that lets you spin up, chain and monitor any open-source LLM on your own GPU (or cheap cloud boxes) in minutes. You get sub-150 ms latency, full data sovereignty and a trace viewer that exposes every token’s cost, heat and hallucination rate—so you finally know what’s happening under the hood. Why you should care: 🤯...

(𝘈𝘯𝘥 𝘯𝘰, 𝘪𝘵’𝘴 𝘯𝘰𝘵 𝘫𝘶𝘴𝘵 𝘢𝘣𝘰𝘶𝘵 𝘵𝘰𝘬𝘦𝘯 𝘱𝘳𝘪𝘤𝘦𝘴) We all look at the per-token cost of GPT and think “meh, a few cents.” But over time, those cents metastasize into ops chaos and architecture debt: 🔻 𝗟𝗮𝘁𝗲𝗻𝗰𝘆 𝘁𝗮𝘅 Every API call across the wire adds 300–800 ms. Stack 3 agents together? Welcome to UX hell. 🔻 𝗖𝗼𝘀𝘁 𝗰𝗿𝗲𝗲𝗽 Hit ~8 million tokens/month and your API bill surpasses the cost of running your own...

🚧 Web App here : https://orka-ui.web.app/ 📦 OSS here: https://github.com/marcosomma/orka 📦 PIP here: https://pypi.org/project/orka-reasoning/ 📖 Docs here: https://github.com/marcosomma/orka/blob/master/docs/OrKa_UI_Getting_Started_With_Images.md 💬 Feedback welcome.

🚀 Features > Modular Agent Orchestration: Define and manage agents using intuitive YAML configurations. > Configurable Reasoning Paths: Utilize Redis streams to set up dynamic reasoning workflows. > Comprehensive Logging: Record and trace every step of the reasoning process for transparency. > Built-in Integrations: Support for OpenAI agents, web search functionalities, and validation...