Comments URL: https://news.ycombinator.com/item?id=46435422

Points: 2

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=45930896

Points: 1

# Comments: 1

So I’m experimenting with a universal booking API: a single JSON interface that lets developers book, cancel, and check availability across many systems (starting with Vagaro, Mindbody, and Fresha).

Under the hood it’s a mix of native APIs and AI-assisted browser automation that handles login states, 2FA, and DOM drift. The goal is to make booking as portable as payments (think Stripe, but for scheduling).

Right now I’m validating demand. If you’ve ever had to build or maintain integrations with scheduling software, I’d love your feedback. What would make an API like this genuinely useful (or trustworthy) for you?

Comments URL: https://news.ycombinator.com/item?id=45765033

Points: 1

# Comments: 1

These are usually purchased by the business owner to run their core business then they but add on features for marketing, review management, etc.

So far, I've just built integrations as they come, but it's definitely hard to scale. Even with AI coding tools, there's a lot to keep track of.

Have any of you found a solution? I'm particularly looking at service booking and appointment scheduling, think hair salons, med spas, and HVAC providers.

I guess I'm looking for something like a Twilio or Plaid that provides a clean API on top of a bunch of messy integrations. Is this out there? Is it worth building if it isn't?

Comments URL: https://news.ycombinator.com/item?id=45754496

Points: 1

# Comments: 2

Coming back a year later and maybe a little wiser, I decided to take another stab at it, but this time I decided I'd build my own integration. I started using Playwright to create a browser automation that allowed my AI agent to book directly in a business's calendar. No integration required!

We're still testing the waters to see if this is received favorably by business owners. Particularly in platforms where they're locked in without needed integrations.

But I also realize this must be a ripe area for engineers to be building, but honestly I don't see a lot out there yet. Despite the maturity of browser automation tech, why aren't there more AI agents connecting to websites? I think what I've build could be useful to other engineers who want to connect their agents to proprietary systems, but don't want to manage the browser automation themselves.

Am I fooling myself? Is there a good reason why this isn't mainstream?

Comments URL: https://news.ycombinator.com/item?id=45726846

Points: 2

# Comments: 1

It took about 9 months of trial and error to get to the right idea, and 3 months more to build the first version of Callio. I’d love your feedback!

# What does Callio do?

Callio is a VoIP softphone with an AI assistant that listens to your calls in real time and suggests what to say next. We believe it is especially helpful for sales calls, and that is who we are marketing it to right now.

You can customize the AI suggetions by:

- Uploading documents (e.g. product sheets, pricing, objection handling guides)

- Setting your sales goals and common Q&A

You can also practice with simulated calls before real calls. The AI uses your materials to generate relevant suggestions that are usable verbatim.

# How does it work?

Connecting an LLM to a call in real-time was hard. We leveraged a lot of off-the-shelf technologies, but wiring them all together was difficult and took some thoughtful effort.

- WebRTC through Twilio facilitates the VOIP call experience via their React Native library.

- WebSockets enable us to stream audio to our backend and the recommendations to the mobile client.

- The LLM gets live transcripts and returns recommendations as the call progresses. We are, in principle, agnostic to the LLM provider, though we've built around OpenAI's APIs to start.

- We built an incremental recommendation workflow to ensure low latency for the AI suggestions despite the latency involved in the REST API.

The architecture is intentionally lean:

- React Native mobile app

- Media server + data/signal server

- FastAPI (initially), but we migrated parts to Sanic for performance

We can certainly do better. Our backend database and some of our data APIs are unnecessarily tricky to scale. Now that we're live on the APP store, we're rebuilding carefully to enable us to scale. Getting real-time to work well (esp. async and latency) was a huge learning curve. Happy to dive deeper if you’re curious.

# What’s next?

The app is live, and we’re marketing to attract more users. Our focus now is:

- Growing beyond our initial beta testers

- Improving the recommendation quality and in-app experience

- Building better onboarding for non-technical users

Would love your thoughts or questions, especially around architecture, user experience, or distribution strategy. The app does require a sign-up, which I know makes it harder for you to try out. Luckily, we have a free trial that you can use.

Thanks in advance for all your feedback!

Comments URL: https://news.ycombinator.com/item?id=44571789

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=24094845

Points: 31

# Comments: 11

Comments URL: https://news.ycombinator.com/item?id=22949677

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=22892270

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=22088836

Points: 1

# Comments: 0