Hacker News new | past | comments | ask | show | jobs | submit | alertify's comments login

At https://bot9.ai (most advanced AI based Customer service engine) we are currently building integrations for CRMs, Ticketing tools, Chat softwares, Ecommerce platforms etc which the bot connects to.

For example the bot connects to freshdesk or intercom and handles the tickets / chats just like a normal agent, and handsover to a human if it can't solve or a customer asks, without the company having to change their chat or ticket tools.

Or connects to shopify to help customers answer any pre-sales questions or help them make a purchase decision.

I see the pain in building these integration one by one and currently exploring a few SaaS that will help me have lot more integrations without us having to build & test one by one (Months).

Paragon , hotglue for example.

There are generally two approaches to do this, each with their pros and cons.

A few go with a unified api approach. The benefit is that you can integrate one Unified CRM API and then use all CRM softwares directly. While this does give a lot of headstart, the downside is that all CRM or ticket softwars may not have the same terminology and you may get into problems.

For example, intercom does not really have a ticket type system but still people use it for ticketing. Whlie freshdesk or Zendesk is a proper ticket system with lots of variables that can be mapped to a ticket. With a unified API you kind of lose out on max potential of the platform you are integrating with, at the cost of faster integration.

Then there are direct APIs which still allow faster integration by handling auth, logging, debugging and some kind of unification but you have to still integrate them one by one in your SaaS. So you gain lot more control, at the expense of time to implement.

By looking at the docs, I see they are trying to go with a hybrid model where they give a direct integration option but i can map the models myself which kind of covers both cases ? Someone from the team can comment. But Interesting.

This is definitely an interesting space which has been mostly covered by zapier. but zaps are ugly, a glue and very expensive for anything serios and mostly used by individuals who just want to connect a few systems together.

In this AI era where everyone is building something with AI and finding out that building the MVP is pretty easy but then gets stuck on expanding because customers wants things to work inside the tools they already use, there is definitely a market for this.

I have sent a meeting request, happy to connect and see if this helps us.


Thank you for the insights and looking forward to chatting!


These are mostly VPS or single server management tools.

Hoy is targeting mid sized compaines looking to build a modern, highly available private cloud platform with distributed compute storage and networking.

- Load balancing across multiple servers

- Auto scaling instances as well as Nodes on demand.

- Fully managed Database & Replicas.

- S3 comptable storage

- Built in observability

- First class dev experience.


So something in the realm of Proxmox or PVC or xcp-ng?

I would agree that the three things I can list here are a hard sell compared to any integrated cloud provider - they all focus on VMs and the rest is up to you, or so my experience has gone so far. Looking forward to trying this out!


Thank you.

A free tier for hobbyists and then small monthly subscription based on number of servers managed.

Addons which are not directly handled by your compute will be charged as well. for example DNS, CDN, Backups etc.


Hoy offers a cloud-like management experience for baremetal servers, aimed at delivering the same ease and flexibility you'd find with services like Digital Ocean and Vercel but for your own hardware.

Baremetal servers can be significantly more cost-effective and performant compared to traditional cloud services, and they often come with generous bandwidth quotas.

The tool supports servers from any provider, including Equinix and Hetzner.

Here's what you can do once your servers are connected:

- Deploy Node.js and Python applications

- Set up static websites

- Manage databases such as PostgreSQL, MongoDB, and Redis

- Install open source stacks like Wordpress, Elasticsearch, and RabbitMQ

For developers, it has automatic git deployments creating a Vercel/Netlify-like environment on your own hardware.

With hoy cli, deploying an app will be as easy as "hoy deploy"

Using bare metal servers means very low predictable & fixed costs, better performance and eliminates surprise billings.

Hoy is aimed towards large cloud like setups offering distributed compute storage and networking on your own hardware instead of being a VPS management system (Like caprover etc).

You can try it out now with our demo, no signup required: https://demo.hoy.sh/

Happy to answer any questions.


I didn't see any mention of JVM apps, I would expect this to be a reasonably large opportunity. With Spring, Quarkus, Ktor, http4k and many more, there are lots of people choosing to build backends on the JVM but deployment is problematic. Vendors like fly.io don't run JVM apps unless Dockerized. They don't work great with AWS Lambda or serverless unless built with Graal. I think lots of developers would like to simply drop a JAR file on a server somewhere and have it run, but I'm not aware of any cloud services which enable this.

I am personally working through deploying on bare metal cloud servers using Ansible and yes, it is challenging to build a full-fledged server with a reverse proxy, local data store, OpenTelelmetry collector, and my JVM app backend -- then secure the server, share keys, run all the services, export logs, open an entry points for a CI deployment, manage env variables, etc. Dokku is an option but has its limits.

The benefit, of course, is that I can rent bare VPS's with redundancy for a year for less $ than most managed services charge per month. If your service made JVM deployments easy while keeping costs low, you might have this market all to yourself.


Dokku Maintainer here.

Curious what limits you run up against with Dokku, and how we might better handle those in the project (whether those are JVM related or otherwise).


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: