After four years of development, we are happy to share Airy with you.
Airy is an Open Source Conversational Platform to store, structure and utilize conversational data in a secure and privacy-compliant way.
With Airy, you can integrate with Conversational AI like Rasa to train smarter models based on actual conversations.
You can host your own open source messaging API to enable your developers to build conversational experiences even for privacy-sensitive industries, such as banking, insurance or healthcare. Airy's core platform is fully open source and runs in your own cloud or even on premise.
We built Airy on Apache Kafka for ultimate scalability, so you can ingest and stream all kinds of conversational data to:
unify your messaging channels
include human agents via an Inbox UI
gain insights from Conversational Analytics
Airy has connectors for conversational sources such as:
Facebook Messenger & Instagram
Google's Business Messages
WhatsApp Business API
SMS (via Twilio)
Airy Open Source Chat Plugin
Custom sources
i have built a similiar thing in the past, called Dialogflow Gateway (https://dialogflow.cloud.ushakov.co), which connects Dialogflow to Web and E-Mail protocols, also open-source
Hi, looks very promising. I would totally consider switching to this streamlined approach to business conversations. However, I don't see any mention of email. Email is IMHO vital for any service that wishes to 'integrate' the full company conversation stack. Do you plan on adding additional integrations?
This is a great point! Email is and will remain _the_ messaging use case for any business. I've created a ticket for adding email as a messaging source so you can track the progress: https://github.com/airyhq/airy/issues/1953
And yes we do plan on adding more sources and are therefore listening to the community to learn which are most in-demand. So thank you very much for the feedback!
Yeah, maybe the word "free" in "Free Open Source
Conversational Infrastructure with Apache 2.0 license" isn't prominent enough. ;)
For Airy Enterprise and Managed Cloud, we usually like to listen to a potential customer's use case first and come up with a custom pricing that makes sense for both sides, usually containing fixed licensing options, volume-based components or even location-based pricing which can make a lot of sense for multi-location enterprises, but rarely works for e-commerce companies.
hey artificialLimbs! Yes, a Teams integration is on the roadmap, as we plan to support all conversational channels and our data model already supports it.
We are working our way through potential channels as we speak, the Teams API was in beta until recently.
We love Matrix for their big vision to build an open network for secure, decentralized communication.
Airy has a more centralized approach from the perspective of a single organization. We want to give organizations an easy way to access all conversational data across their entire organization in a structured form and help them to utilize it, e.g. to train smarter machine learning models in the interest of their customers.
This is reflected in our mission of structuring the world's conversational data to power the future of customer experiences.
A bit of advice. When someone clicks on "Pricing" there should be information involving dollars (or another currency depending on where you're located).
For certain products it can make sense to only provide services on an as-needed basis with pricing tailored to how much you think it'll cost to run the service for that customer plus some margin. For this I think it's less like that (seems like a Zapier clone for chat, based on my initial impression).
In my previous job I was the lead engineer of the MessengerPeople Unified Messaging API, so congrats, this looks awesome! :)
I wonder how you solved media files and attachments? I remember having major headaches around transporting, storing and retrieving media files for different messaging backends, especially WhatsApp and Telegram.
Also, any plans to support WhatsApp Business containers directly?
Yes this was a very challenging problem. After many iterations we came to the conclusion that in general it's best to have no opinion on the structure of the content that is being sent. So we store every message exactly as we receive it and do platform specific file interactions in a separate stream. We then store the results in a metadata topic and use that to render messages on-demand. This immutable approach makes retrying and storage migrations easy and safe.
Regarding WhatsApp Business containers we are still waiting for them to wrap up their Beta. Currently we support WhatsApp and SMS via our Twilio messaging source (https://airy.co/docs/core/sources/whatsapp-twilio), but we definitely plan on making it a first class citizen.
Hey, congrats on the launch! it seems really nice and helpful. I was thinking of creating something similar for the communication channels of my SAAS applications.
Is this product for small size businesses? Does it need a single cloud vm or it need several servers and services, a ton of configuration and an IT degree to manage it?
In general, Airy is mostly targeted at mid market companies and enterprises that deal with a lot of conversations, e.g. we helped a European retail company to launch conversational experiences on Google Search and Google Maps for their central customer support team and their 1,800 local stores (https://businessmessages.google/success-stories/tedi/).
Airy gives you an enterprise-grade communication infrastructure, running in a Kubernetes cluster, for example EKS with several virtual machines when you are running on AWS. Our recommended initial setup consists out of two c5.xlarge EC2 instances. This should be powerful enough to handle several conversational sources and a few hundred thousand conversations per month. We also have a few rather large SMBs among our customers, but the average SMB rarely gets to such amounts of conversational traffic yet.
Installing Airy is rather easy and can be done with our Airy CLI to set up a remote cloud instance from your local machine with a single command e.g. in AWS ("airy create --provider=aws").
How does Airy compare to Jovo (https://github.com/jovotech/jovo-framework) other than supporting different channels (Jovo seems more focused on voice whilst airy on text?)?
According to my understanding, Jovo seems to be a rather light-weigh framework based on Javascript/Typescript to build voice experiences e.g. for Alexa, Google Assistant, etc. I have not tried it out yet, but it seems to be a cool project to build these kinds of experiences primarily around voice.
In comparison, Airy is a much more resource-intense backend service running conversational infrastructure that you couldn't run on comparable hardware. Airy is designed to run in the cloud giving you a Kubernetes cluster with all the components to stream conversational data at enterprise-scale. For now, we primarily focus on text-based communication, but could theoretically also support transcription and processing of voice based messages or live videos.
Congrats on the launch.
We have been using Chatwoot (https://chatwoot.com/) - an open source solution for customer messaging. Great to see all the open source project. How are you differentiating Airy?
Thank you for the question, soorajchandran. We appreciate to see more open source tools in this space and believe that Chatwoot is a great choice when you are primarily looking for a cost efficient alternative to Intercom, Zendesk or basic contact center software with a UI for customer support agents.
What you get with Airy is enterprise-scale conversational infrastructure that can power millions of conversations simultaneously by ingesting messaging events in Apache Kafka, running in a dedicated Kubernetes cluster to stream and process conversational data for a variety of use cases, such as integrating with Conversational AI platforms or storing all your conversations in a data lake to run conversational analytics or train machine learning models based on actual conversations.
In that sense what we do is more like "Segment" for conversational data.
In our approach, we would rather like to integrate with for example live chat plugins provided by Intercom instead of replacing them at companies that already chose Intercom to serve their customers with a live chat plugin on their website or in their mobile apps. Airy also comes with a fully customizable open source live chat plugin and an Inbox UI for human agents, but it's not at the core of what we do.
We believe there is much more value to be gained from utilizing conversational data and we therefore like to integrate and play well with other solutions in the space as we believe that companies should have the freedom to choose the tech stack that best suits their requirements and budget restrictions.
With every messaging and conversational product I fear that the pricing is volume based, the more conversations, the more you pay. How does Airy handle pricing?
Thank you for that question, michaelagustin. You are right: most popular SaaS tools in the conversational space are indeed priced per volume, usually per message or per (active) conversation.
As communication is core to every business, we strongly believe every company should own their conversations and utilize their conversational data in the best possible way, taking the interest and privacy of their customers into consideration.
We believe there is a unique opportunity now in the market to create an open source conversational stack and we would like to contribute to it with something we are good at and have a lot of experience in. Our goal is to create an open standard for the processing and storing of conversational data which is why we went open source.
This situation is perfectly suited from our perspective for an open core pricing model where we will continue to give away our Airy open source core platform for free under an Apache 2.0 license and sell additional enterprise licenses for optional features like advanced routing capabilities, team management, advanced storage and analytics solutions on top to enterprises that have additional requirements and more organizational complexity to deal with. Enterprises can run Airy Core + Airy Enterprise in their own private cloud or even on premise for privacy sensitive industries like banking, insurance or healthcare.
For business teams that want the full power of a conversational platform like Airy, but can't or don't want to dedicate engineering resources on their end, we also offer a fully managed Airy Cloud solution on the side. Because each Airy instance is fully independent, we can even offer this service in any region in case the relevant customer has preferences e.g. to store their conversational data exclusively in the EU or in a specific country or data center.
Airy is a conversational platform, built mostly for businesses: most enterprises have a variety of conversational apps and channels they support (from Facebook Messenger & Instagram for Customer Service to their own livechat for sales, etc). Airy helps these businesses bundle these channels, store the conversations and power the different usecases.
It looks like Mixin is an open source cryptocurrency wallet that also has peer to peer chat and a desktop version. So the only common points I see is that both projects are Open Source and use chat as an interface.
Hi renrutal, Steffen from Airy here. Thanks for your remark and I couldn't agree more with you which is why we don't pay-wall privacy features. Actually, nothing stops you from turning on encryption also in the open source version of Airy as Pascal pointed out above.
The topic you mentioned above comes from our pricing page for the Airy Enterprise Edition and is about an additional (!) and fully optional conversational data store for archiving conversations and to provide for conversational analytics use cases. We currently only offer this additional streaming option for enterprise customers with large amounts of conversational data by leveraging data lakes on economic cloud storage solutions like AWS S3. We strongly recommend to activate server-side encryption for this storage option.
Hi renrutal,
Pascal from Airy here. Encryption is a very important topic for us. Since in all our versions all of the data resides in Apache Kafka and is only exposed via Kafka Streams Apps we are able to take advantage of Kafkas SSL encryption. For encryption-at-rest we use our cloud providers disk encryption.
We also support Open ID Connect as an identity layer so we can focus on improving our platform and be assured that our authentication is always up to date and secure.
You raise a good point though and we will continue innovating to make encryption accessible to everyone.
Is that E2E, though? I'm not Kafka-knowledgeable, but that sounds like it's encrypted as it gets to your servers and encrypted as it rests on them -- not encrypted all the way between people sending messages.
I'd agree in general for consumer use-cases but not for business use-cases. With an open source business tool that may be used in non-business context, charging for features required for business use makes sense.
Even among consumer use-cases, lack of privacy may be a feature. In spatial chat, for example, being in able to overhear conversation within 'earshot' is a feature. Selling private space in that context makes sense and similar to selling improved voice quality, at least to me, and there are stark operational cost boundaries in the involved tech that can complicates the picture.
> We live in a world where companies and governments are actively spying and harming those under them.
Therefore this random company owes you security features for free?
Users wanting free shit is the reason why companies like Google and Facebook are doing all this spying in the first place. Those huge warehouse-size datacenters don't pay for themselves.
After four years of development, we are happy to share Airy with you.
Airy is an Open Source Conversational Platform to store, structure and utilize conversational data in a secure and privacy-compliant way.
With Airy, you can integrate with Conversational AI like Rasa to train smarter models based on actual conversations.
You can host your own open source messaging API to enable your developers to build conversational experiences even for privacy-sensitive industries, such as banking, insurance or healthcare. Airy's core platform is fully open source and runs in your own cloud or even on premise.
We built Airy on Apache Kafka for ultimate scalability, so you can ingest and stream all kinds of conversational data to:
unify your messaging channels include human agents via an Inbox UI gain insights from Conversational Analytics
Airy has connectors for conversational sources such as:
Facebook Messenger & Instagram Google's Business Messages WhatsApp Business API SMS (via Twilio) Airy Open Source Chat Plugin Custom sources
Check out a short intro video of Airy here: https://youtu.be/zwDosYHitYg
You can start trying it out by reading on our website: https://airy.co/ph
If you like what we are doing, please give us a star on Github: https://github.com/airyhq/airy
And we are of course happy to answer your questions!