Thoughtly Community Icon

Join our Slack Community

Thoughtly Virtual Agent v1.5: New Features and Improvements Released! | Thoughtly Community

Thoughtly Virtual Agent v1.5: New Features and Improvements Released!

February 11, 2025

Reading time: 2 minutes

Thoughtly Virtual Agent v1.5: New Features and Improvements Released!

v1.5 is finally here 🎉 🎉 @channel the revolutionary new Thoughtly Virtual Agent, v1.5, is now available. Thanks to everyone for your patience while we rolled this out. This huge update includes a host of new features and improvements, including:

Mid-call Actions: Agents can now perform actions during a call, such as sending an SMS, updating a CRM, or fetching real-time data.
Reduced Hallucinations: Thanks to a proprietary combination of mathematical calculations and LLM reasoning, agents now generate more accurate responses with optimized context windows.
Improved Failure Tolerance: The system now leverages multiple LLM vendors to enhance reliability and minimize downtime.
Lower Latency: More efficient logic and scalable infrastructure reduce response times for a smoother user experience.
Enhanced Navigation: Edges now carry meaning instead of relying on bloated context, leading to smarter conversation flows.
Advanced Node Controls: Users now have more granular control over settings, including:

Voice Confidence Threshold – Adjust response certainty.
Sensitivity Threshold – Fine-tune agent responsiveness.
Verbatim Mode – Force strict adherence to scripted responses.
Disable Interruption – Prevent the agent from speaking over the user.
Small Talk & Music Handling – Configure how the agent responds to casual conversation and background music.

Cartesia Voice Support: Expanded capabilities with Cartesia for hyper-realistic, low-latency voices.

Try it out by enabling experimental features in your Developer Settings. Be sure to follow our Tips & Tricks to prompt correctly and get the most out of the new Agent. Read more in our Changelog 🙌

22 comments

Mark Garrett
·
·
Vre6news Torrey Leonard, do we have any training videos on these updates?
Torrey Leonard
·
·
Coming soon Mark Garrett! Since this is a beta release, we're opening things up for everyone to test. Once the testing has gone well and the UI is complete, we'll do a whole training set
Mark Garrett
·
·
Thanks Torrey Leonard. So to confirm, this is beta and not production? What happens to existing thoughts? And do we have the option to build in 1.0 production or 1.5 beta?
Michael Smith
·
·
Starting a new agent seems a little different to me. Marketplace is greyed out and it seems like more dropdowns. Corrected to say agent.
Torrey Leonard
·
·
Correct, this is still beta. Existing agents will remain the same, but we'll have an ability to "upgrade" them once we're out of beta. For now, you can choose either v1 or v1.5 when creating a new agent (if you have experimental features enabled)
Torrey Leonard
·
·
Great callout Michael Smith, yes the create modal looks new as well. This should be helpful Mark Garrett 👍
Mark Garrett
·
·
Well done!
Mark Garrett
·
·
Thanks for getting this out lots of features that we've been waiting for. Excited to play with this!
🙌2
Vinícius Vicente Alves Terçariol
·
·
Hello! Great! I’m testing the new version, the nodes speak don’t accept variables, only in the first step (start). I try add @ to list the variables, but don’t show any variables. Thanks Torrey Leonard
Ryan Prince
·
·
Torrey Leonard In the previous version, I could loop a speak node to itself when the customer had more questions, but now it won't allow me to do that. How can we work around this?
Ryan Prince
·
·
Vinícius Vicente Alves Terçariol
·
·
The voices of Cartesia are great!
🙌1
Ida Ma
·
·
Good news! Thanks. Just to confirm, we don't need to do anything with our existing Thoughtly, is that right?
Vinícius Vicente Alves Terçariol
·
·
Any updates?
Torrey Leonard
·
·
Hey Ryan Prince, great question. We'll have the "Objective Node" out in the next week or so for v1.5. That'll take care of the use case of looping a node back to itself. Ida Ma, this is a different type of agent, so you'd have to create a new one. For context, the "v1.5" agent is more linear, less conversational, so if you have an inbound "IVR" style use case, this is best. If you have an outbound "sales-y" type of agent, then v1 is better. Soon, both v1 and v1.5 will have the exact same features, but they'll be labeled as two different models: Model 1: v2-conversational (the standard version you're used to) Model 2: v2-linear (this version)
💯1
🙌1
Torrey Leonard
·
·
Vinícius Vicente Alves Terçariol can you try variables now? it should work 👀 If not, you can always use { variable }
❤️1
Vinícius Vicente Alves Terçariol
·
·
Yes, I try now! Let me check!
Vinícius Vicente Alves Terçariol
·
·
I try with the many and one variable, the builder are ok, but in the call no, the variable in the call, not change to value.
Adam Aktar
·
·
🔒[private user] morning
Shay R
·
·
How soon is soon to get mid-call actions in v1 Torrey Leonard? v1 and v1.5 feature parity that is. Lmk! Cheers!
Eetu Häkkinen
·
·
heyhey - I noticed it's having issues with GHL. Giving permission errors when turning on v1.5 - can't load all subaccounts and/or all calendars from specific subaccounts. Everything works normally via v1 -- Bryan Alagao has been messaging you guys Torrey Leonard about this for few weeks now. I looked into it yesterday and found out the v1.5 being the issue So, when do we get fix to that one? v1 works for now tho
Eetu Häkkinen
·
·
v1.5 works when connecting just one subaccount instead of agency v1 works when connecting agency as well