Free and fully open source (MIT)
Yacana is designed for both beginners and advanced AI users. It features a simple OOP API with a smooth learning curve, while also offering advanced runtime LLM configurations when needed. The real strength of the framework lies in its ability to deliver impressive results with open-source models, even small ones, making tool calling effortless with any LLM. For production-grade capabilities, Yacana offers a guided workflow approach or multi-turn chat, both leveraging what we typically call 'Agents'. However, Yacana takes a different approach to Agents compared to other frameworks, focusing more on chaining Tasks together rather than on the Agents themselves.
Like LangGraph, Yacana has the ability to guide any LLM in a specific direction. Open source LLMs are limited compared to frontier models, which is why you need to chart a course they can follow to produce accurate outputs. Yacana helps you achieve this without relying on hard-coded links or static configuration files—just pure object-oriented programming (OOP).
Like CrewAI, Yacana enables LLMs to communicate with each other, allowing them to brainstorm while keeping control. Yacana provides various methods to stop the conversation at the right moment, ensuring it produces a computable and useful result.
Tool calling is the ability to invoke 'classic' programming functions, allowing an LLM to interact with any software component. Currently, other frameworks only support tool calling on fine-tuned LLMs specifically designed for this feature. However, Yacana enables tool calling with any LLM, even those that don't officially support it. This means it works with even the smallest models, like PHI, making Yacana ideal for IoT applications.
Entering the AI world can be challenging. That's why Yacana was designed with a clean and simple programming API to help prevent developer burnout—or throwing their (expensive) computer through the window. We've ensured the documentation is always up to date and includes plenty of examples. Please note, however, that Yacana has currently only been tested with Ollama, but we are working on integrating other inference servers in the next update.
It's new, it's shiny, and most importantly, it's simple! Get started and have fun developing with AI right now on your very own computer.
Yacana comes from the Inca mythology. It's a constellation known as the kamac (spirit creator) of the llamas, that is, its vital force, the soul that makes them live.
Use natural language to interact with your K8s cluster. Gno6 will translate your queries into valid Kubectl commands. A real-world use case from which you can get inspiration.
A great article about why Agentic frameworks are hard to use and how Yacana can integrate in YOUR software applications with ease!
Nothing to show here... Yet!
Nothing to show here... Yet!
Video tutorials to help you use Yacana efficiently!
In need of help ? Ask
here
Found a bug ? Report it here