Subscribe
Sign in
fixa
p/fixa-2open-source voice agent testing and observability
Start new thread
Garry Tan

fixa - open-source testing and observability for voice agents

by•

fixa helps you run simulated tests, analyze production calls, and fix bugs in your voice agents. oh, and we're fully open source.

get started for free - no demos, no commitments, only pay for what you use.

Add a comment

Replies

Best
Jonathan Liu
Maker
📌
hey there 👉👈 this is my second launch of an open source product here on product hunt. and this time it’s in voice agents. 🗣️ @oliver_wb and I were both voice agent developers at PlayHT – and, in order to find bugs in our agents, we would spend countless hours manually calling them and listening to conversation recordings. so, we’ve decided to build an open-source platform that does this automatically. how does it work? 🤔 1. run tests 📝 our voice agents will call your voice agents to test them out. you can run tests via github actions, API call, or directly in the UI. pinpoint exactly where errors occur in the audio along with an analysis of what occurred. 2. monitor production calls 📞 we help you identify latency, interruptions, and correctness of your calls in real-time. set up slack alerts for calls where latency goes above a certain amount or when the agent fails to collect the user’s email address, etc. check out our website at https://www.fixa.dev get started for free - no demos, no commitments, only pay for what you use. and make sure to star our github - https://github.com/fixadev/fixa let us know if you have any feedback in the comments below!
André J
I recently chatted with the slack support. I felt it was a bot, because it could never get anywhere. Just kept the convo going without resolution. Is there a way to test resolution ability? I guess thats when you ask for a real human?
Jonathan Liu
Maker
@sentry_co fixa is primarily focused on voice agents for now, not chat bots, but yes there could be a way to test resolution ability by running a bunch of test cases with the bot and analyzing chats in production where a resolution was not reached. in your case, I would recommend that you ask to talk to a real human 😭
André J
@jony1266 The emails Kept replying with Real first names. So I didn 't want to offend anyone by indirectly calling them a robot 😅. I got out of the infinite loop by replying with a proper turing test question. I think their mainframe blew up after that and I never heared back. I guess when it comes online again it will continue 😅
Jonathan Liu
Maker
@sentry_co 🤣🤣🤣 oh my
Sam @CRANQ
Wow, what's been the choice to allow it to be open-source, apologies if that's not a smart question but I'm intrigued :) I guess we'll be able to achieve a bunch more now that you guys have made this available though!!
Jonathan Liu
Maker
@cranqnow not a dumb question at all! we decided to go open source for 3 reasons: 1. build developer trust - for a dev tool, we believe that transparency is key, and being open source allows us to do just that. this also helps with distribution, with people recommending us to their friends, if they trust us and love the product. developers, being highly technical, can also self host if they prefer (especially important for enterprises who need on prem deployments). 2. handling sensitive data - we're processing thousands of user call recordings a day. allowing you to inspect how exactly a call is processed gives you peace of mind that we aren't doing anything sus with your calls. 3. testing / observability should be accessible to all - we believe that testing and observing your calls is a crucial part of the development process, and anybody developing a voice agent should be able to use these tools regardless of budget.
Sam @CRANQ
@jony1266 Ahhhh ok this makes a lot more sense to me! Thank you very much for making it clear!! Of course there’s a lot to be said for getting kudos for contributing to the community & not necessarily asking for anything in return etc. Naturally the sensitive data part is very clear too! Wishing you the best of luck with the rest of the launch!! I appreciate you taking the time to help me understand more!
Idris
Dope product & killer team - you can’t go wrong!
Jonathan Liu
Maker
@idrisyunis_ you know it 😤
Marc @ techire ai
Great product, and opensource🙌
Jonathan Liu
Maker
@marctechire appreciate it, mark!
Tereza Stejskal
It looks like a very promising tool, we will definitely going to try that in our app!
Jonathan Liu
Maker
@tereza_stejskal perfectt, let us know what you think :)
seb
love fixa
Jonathan Liu
Maker
@sebselassie thanks seb 🥺
Andy Li
Fixa is amazing at testing voice agents!
Jonathan Liu
Maker
@andydy25 glad to hear it 🥺
Andrea Juwono
This is groundbreaking! Can't wait to see where else the team takes this 🌟
Jonathan Liu
Maker
@andrea_juwono we're going to take it so far 🚀
Tony Xin
really excited for this product 😸
Jonathan Liu
Maker
@tony_xin me too !
Shivendra Soni
Hey @jony1266 , brilliant product and well timed. I have a question though, how do I test it with my existing voice agents? I would love to test my voice agent as a blackbox where fixa interacts as if its an outside entity not a first class citizen of my infra.
Jonathan Liu
Maker
@shivendra_soni2 that's exactly what we do! you can give us a phone number to call and then our test agents will interact with your voice agent as a black box. all you need to give us is a prompt for how our test agent is supposed to behave.
Mohamed Gaber
Brilliant idea! How many languages do you guys support?
Jonathan Liu
Maker
@mohamed_gaber4 we support all the languages that deepgram supports! check our docs here (esp the language parameter): https://docs.fixa.dev/api-refere...