Hi DEV, I'm Marco. Four months ago I started developing OrKa, a modular orchestration engine defined by YAML for agent reasoning. Today it's not just a prototype: it's a functional system with live memory, explainable flows, and a visual builder, but it's already too much for one person to maintain.
I work solo with a single repository. OrKa has become a cognitive infrastructure stack: structured memory layers, Redis and Kafka queues, traceable agent chains, a visual interface, and measured orchestration loops. Version 0.8.0 works. Even so, it's a struggle and I need help.
Why OrKa exists: ending black-box AI LangChain, Flowise, and AutoGen chain things together and call it cognition, but if you try to trace memory between agents, understand why a step was taken, or observe reasoning in real time, you can't. OrKa is different: modular agents with independent logic and memory, cognition flows defined in YAML instead of tangled scripts, traceable logging in Redis and Kafka for every agent decision, observable interface and terminal TUI, six-layer memory model with decay and scope-based storage, and confidence-weighted routing to simulate dynamic reasoning. This isn't prompt chaining; it's explainable, testable, and local cognitive orchestration.
What works in OrKa 0.8.0 Fork and join execution with dynamic routes, per-agent confidence scores and agreement synthesis, Redis and Kafka compatible logging, visual OrKaUI with YAML synchronization, local and API support, ServiceNodes like RAG MemoryWriter and embedding fetchers, benchmarks: 1000 loops with average latency of 7.6s and approximate cost of 0.00011 cents per run on DeepSeek 32B.
But it has real problems This is truly open source. I'm drowning on several fronts: functional memory scope but v0.8.0 needs a simple fallback using Redis GET and SET; the Orchestrator class is bloated and needs modular separation; UI gaps: memory nodes are templates, trace replay works halfway; outdated documentation and scattered examples in the examples folder; low visibility in communities like Reddit or X; and the hell of being a solo developer writing infrastructure, fixing YAML bugs, answering on Discord, and also raising three kids.
What you can do today Build SimpleMemory in src/memory/simple.py using Redis SET and GET and set aside Kafka overhead, improve fallback logic for small LLMs. Clean up the codebase by separating src/orchestrator.py into smaller modules, add ruff and mypy and apply strict linting, improve logging by differentiating agent and orchestrator. Contribute nodes or agents: add a PlannerAgent, extend RAGNode with support for Pinecone and Chroma, create MemoryVisualizerNode for the UI. Write real documentation: docs/architecture.md, polish examples/fact_checker.yaml, and create quickstart.md for non-engineers. Spread the project: share your flow on X, Reddit, and Discord, publish your own YAML orchestration, and use it in hackathons.
AI shouldn't be a black box. If it can't be traced, it shouldn't inspire trust. OrKa seeks to build cognitive systems that are transparent, local by default, composable, and deterministic. It's not a LangChain clone or a chatbot wrapper: it's a runtime for modular cognition. You define the graph, OrKa executes it, traces it, and explains it.
Practical request: don't just star the repo, run it. Installing is as simple as running pip install orka-reasoning and trying an example YAML file like orchestrator: id: fact_checker strategy: sequential agents: - id: validator type: binary prompt: Is this statement factual
If you want to fix docs, build a node, or just hang out on the server, I'm available. Fork the repository and collaborate. Repository on GitHub: https://github.com/marcosomma/orka-reasoning
About Q2BSTUDIO Q2BSTUDIO is a custom software and application development company specialized in artificial intelligence, cybersecurity, and AWS and Azure cloud services. We offer custom software solutions, custom applications, and AI agent integration for businesses, as well as business intelligence services and Power BI consulting. Our team develops AI solutions for businesses that include custom models, agent orchestration, data pipelines, and cybersecurity practices to protect critical infrastructures. If you're looking for AWS and Azure cloud services, artificial intelligence consulting, and Power BI integration to boost your dashboards and analytics, Q2BSTUDIO can help you design and implement scalable and secure solutions.
Keywords for positioning: custom applications, custom software, artificial intelligence, cybersecurity, aws and azure cloud services, business intelligence services, ai for businesses, AI agents, power bi.
Let's build the cognitive runtime we all want while also driving robust enterprise solutions with Q2BSTUDIO. If you're interested in collaborating on OrKa or exploring professional services, get in touch and contribute to the open source.
