Maintainer note
People have been reaching out to me lately about Orka-reasoning.
I stopped development last month. Not because I lost interest, and not because I changed my mind about orchestration being the real key. I did not.
Orka was never meant to be a product. It was mostly a playground for me.
2025 was a hard year professionally. I saw the weaknesses of AI demos in real life. I saw the lack of scalability, the overuse of single-agent patterns, and often no real agent design at all. The prompt-and-go strategy worked for demos, but it was not scalable.
That is where Orka came from.
My first commit was just a small POC with two agents processing one input in a linear flow. From there, I added routing, branching, loops, and more. I built it in public because I wanted to show that there was a better path.
Some people engaged from the beginning. Others arrived later. Over the year, I saw around 30k downloads across GitHub and PyPI. Some people came back with feedback. Some contributed. Many criticized it. All of that helped.
Orka kept growing and became a sort of personal AI ecosystem for me. A place where I could build agentic flows with no boundaries, and where critical thinking was part of the pipeline before producing a final output.
It was fun.
Then, slowly, I became a maintainer.
The pressure from download growth started creating expectations. I did not want to release new versions without testing properly, checking backward compatibility, and making sure things were stable. That made me slower, and it also tied me to a narrower problem than the one I actually wanted to explore.
Then 2026 arrived, and more people started catching up with the limitations of AI demos. Investors began asking for real results, not just flashy prototypes. Many products had to rewrite themselves completely. They learned, but from a business perspective, many also burned money and are now dealing with expectations that do not match the real cost of AI.
In my view, this is the silent AI pop. The bubble is not exploding loudly. It is slowly deflating and hitting the ground.
So this is the point.
Orka was fun, and I am quite sure it is full of small gold stones if you want to dig. But it was my playground, not a production-ready product, and it will never be.
You are free to fork it, branch it, and build on top of it. But there will be no maintenance from my side.
Overview
OrKA-reasoning is a open-source, local‑first, YAML‑driven system for composing AI workflows. Define agents and control‑flow in configuration, run them with a single command, and keep everything observable and reproducible.
Think of it as a streamlined, open-source experimental framework with a strong focus on:
- YAML‑first workflows: declarative configs over code > see YAML Configuration
- Visual builder and runner (drag‑and‑drop) > see OrKa UI
- Built‑in memory with vector search and decay > see Memory System
- Local and cloud LLM support with cost controls > see Backends
- Rich control flow (router, fork/join, loop, failover, plan validation) > see Agents & Nodes
- Path discovery with GraphScout (beta) > see GraphScout
- Structured JSON inputs for complex data > see JSON Inputs
- Observability and tracing > see Observability
- Testing guidance and examples > see Testing Guide
Get started
- Quickstart: docs/quickstart.md
- Documentation index: docs/index.md
- Examples: examples/
Community
- Issues & feature requests: https://github.com/marcosomma/orka-reasoning/issues
- Contributing guide: CONTRIBUTING.md
License
Apache 2.0 — see LICENSE
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file orka_reasoning-0.9.17.tar.gz.
File metadata
- Download URL: orka_reasoning-0.9.17.tar.gz
- Upload date:
- Size: 814.2 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/6.2.0 CPython/3.12.8
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
2815258bab564cd5be583d153a856978ed8530794b1a0d8b8010e27fd4dc72d6
|
|
| MD5 |
16ed343d7a5aa642b3ec92f4f8a99543
|
|
| BLAKE2b-256 |
66bf6d99309a5e1f7c255385fdf227a705e7b72c7ea6e1819e8ecbfac197db1a
|
File details
Details for the file orka_reasoning-0.9.17-py3-none-any.whl.
File metadata
- Download URL: orka_reasoning-0.9.17-py3-none-any.whl
- Upload date:
- Size: 856.5 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/6.2.0 CPython/3.12.8
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
90fc32c8c15c7dd51277d45b4a8161d9fbeb7f52719966b2203d0efb036d152e
|
|
| MD5 |
3240adc4995ea944b5b843ca03b6ed78
|
|
| BLAKE2b-256 |
945325b9f9c3f267c6c9e6e30c56a84971124ab3f8e98bdfb1131eb7a915fbfe
|
