I use ReAkh to trigger builds across two Jenkins environments. Each one has around a dozen Git parameters, different branch naming rules, and separate version numbers to keep track of. Before, I'd always have to stop and double-check whether a branch needed the origin/ prefix, make sure I wasn't reusing a version number, and verify I was in the right environment.
Now I usually just tell ReAkh which branches I want to build. It handles the Jenkins login flow, CSRF token, parameter formatting, and build request automatically. If something fails because of a temporary network issue, it normally retries on its own, which saves me from watching every request.
One thing I didn't expect is that it keeps track of the workflow over time. For example, once we figured out that one Jenkins instance requires the origin/ prefix while the other doesn't, I haven't needed to explain that again. We've also run into occasional 504 errors from Gitee, and it generally retries without me asking. After using it for a couple of weeks and running around 30 builds, the repetitive parts of the process have mostly disappeared.
Poe
Hey Product Hunt, I’m one of the developers behind ReAkh.
We built ReAkh around a simple pain point: AI chat is great at answering, but real work often needs more than an answer.
ReAkh supports multiple model families, including DeepSeek, Qwen, GLM, Kimi, and MiniMax. But model choice is only one layer. What we care more about is how to turn a model into a useful working environment.
For us, that means:
grounded answers with web search and Knowledge/RAG
custom AI teammates for focused roles and repeatable workflows
sandbox execution for code, calculations, and file generation
downloadable outputs like PDFs and Markdown
the option to connect your own machine for private workloads
What we’re exploring is whether AI chat should become more like a workbench: one place to research, reason, run, verify, and produce artifacts.
Where do you draw the line between a chatbot and an AI workspace you would actually trust for daily work?
We’d especially love feedback on model switching, sandbox safety, source grounding, custom AI roles, and workflows that still feel fragmented today.
Hi Product Hunt! 👋
I'm Jim, the front-end developer behind reakh, and I'm thrilled to share our AI product with you all today.
As a frontend dev, I wanted to truly stress-test what our AI could do. So, we decided on the ultimate dogfooding experiment: we used reakh to build its own official website (www.reakh.ai). Watching our AI generate the structure and code for its own home was an incredible "Inception" moment!
Here is a transparent look at my experience using reakh as my AI co-developer for this task:
The Pros (What went great):
Zero to MVP in Record Time: We went from a blank canvas to a fully structured landing page incredibly fast. It intuitively understood the layout needed for a modern AI product site.
Excellent Context Retention: It held the context of our product features and brand tone perfectly without needing constant prompt reminders.
Vibe to Code Translation: It successfully translated abstract design ideas into concrete HTML/CSS frameworks that looked great out of the box, saving me a ton of boilerplate coding.
The Cons (Where I still needed to step in):
Pixel-Perfect Tweaking: While the foundation was solid, achieving "pixel-perfect" responsive design across mobile and desktop still required some manual CSS adjustments on my end.
Complex Iterations: For a few highly specific interactive UI elements, it took multiple prompt iterations to get the logic exactly how I envisioned it.
I’m using ReAkh to create a custom AI teammate for my Mac mini, and wanted to share the prompt idea I’m testing:
“Act as Pika Agent, a local AI teammate running on my Mac mini. At the start of each session, load my AGENTS.md as the global rule set, load available skills from ~/.reakh/skills, and use them as working context. Help with local file tasks, terminal workflows, and project-specific automation. Follow my local rules first, ask for clarification when needed, and require explicit approval before sensitive file changes, commands, or actions that may affect system security or data integrity.”
This is the workflow that makes ReAkh feel different for me: instead of a generic chatbot, I can create a teammate that understands my own environment, follows my rules, and reuses my local skills.