orkes 0.1.2

Lightweight multi-agent orchestration framework

Orkes

🔍 What is Orkes?

Orkes helps you coordinate LLM agents using plain Python. No magic, no unnecessary layers. Just explicit control flow, transparent agent logic, and simple message passing.

📝 Background

A while back, I was tasked with ensuring that our agentic-based streaming calls closed properly when a client disconnected.

Out of the box, there wasn’t a straightforward way to close the underlying HTTP connection. Why does this matter?

Because in our case, leaving connections hanging was detrimental to keeping our self-hosted LLM stable and reliable.

The frustrating part: today’s higher level libraries are abstractions on top of abstractions, hidden under even more abstractions, and at the end hidden in dependencies abstraction, layered until a simple fix turns into a complete clusterfuck.

Hence the pain of using high-level abstraction frameworks, some niche cases just don’t get covered.

📚 Documentation

• orkes.agents – Documentation for agent classes, interfaces, and tool integration.
• orkes.graph – Guides for building, compiling, and running DCG-based workflows with OrkesGraph.
• orkes.services – Details on LLM connections, prompt handling, and response parsing services.

🔹 Core Principles

• Explicit control flow — use DCGs, FSMs, or plain loops
• Transparent agents — define prompt, tool, and logic directly
• Simple message passing — plain dicts, no graph state magic
• Minimal dependencies — only what you truly need
• 100% Pythonic — easy to read, modify, and extend
• Stateless by default — you control memory and state
• Hackable and debuggable — nothing hidden

🛠️ Project Status

This is the initial stage of Orkes.

• Vision and core philosophy
• Elementary Directed Cyclic Graph orchestration engine
• Agent structure (prompt + tool + fn)
• LLM related interfaces
• Minimal examples
• Documentations

License

This poject is available as open source under the terms of the MIT License.