Ok so looks like the complex model escalation does need to return to preserve tokens! Will fix in the coming days! Really fine balancing act 
1 Like
If we could add our own API to the tool, it would solve the issue 
1 Like
It would but then I’m on the hook for overspending your tokens not just my own. I’ll put it on the list! Need to then ensure we use models and APIs you bring - which provider do you use? I’ll begin there!
1 Like
I currently use ChatGPT.
1 Like
Me too (paid chatgpt…)
This might have some interesting info and thoughts😉
1 Like
Thanks ppl that’s super great, I’ll check out the pod.
New deploy which is way more frugal and ‘just as clever’ is out - I had my model router sorted just now (was turned off all day hence burning credits).
Let’s see - if you get a ‘blank response’ please start a new chat. Old chats may be unreliable.
Consuming as much info as I can on getting them to behave and thinking I’m nearly there. Using the Gemini family - they are well trained on fitness data (Google use them for their fitness stuff I believe). Stretch goal would be finding some data and training models, but I think we can make something very useful before we’re there!!!
Have a good Sunday all 
Thanks for sharing your app.
I’m still struggling to get it to view my intervals hrv data. It’s from Apple Watch sync via HealthFit. Keeps saying the field is blank despite me seen it in intervals.
1 Like
I may be on my own with this view, haven’t looked at the product but, is it not a bit ‘cheeky’ (being polite) to use a name for your product, which is the brand from which you take the data - then call it ‘Pro’, then charge people to use it?
I can see all sorts of IPR, copyright, trademark topics with calling your product ‘Intervals Pro’.
1 Like
@NotSoTurboTippy it’s a great point and I wouldn’t be doing it if I hadn’t spoken with @david and first sought his approval!
We are totally dependant on (and always will be) the intervals icu platform. Therefore we can drive custom - we plan to encourage supporter subscription, and run our own marketing campaigns so should be beneficial for icu.
Give the tool a try!
1 Like
Thank you for clarifying @Bigblob , I was getting twitchy for a minute 
I’ll take a look.
1 Like
not sure if its user error by me but every request ends up hanging (this has been waiting for 1 hour plus now)
1 Like
Hey @sandeater - please hard refresh the site and try a new chat. I’m not seeing any issues of concern - try a simple question I guess, like ‘what’s my ftp?’ to test!
DM me if you have further issues!
When you start thinking too much and can’t get out of it. Refresh the page and it will return to where it was. I hope this helps!
1 Like
thanks, I tried a few times, but it always hangs on “waiting for approval”. How is this made it is a client or server side request? Does this go to any strange url (just wondering if I have any strange config on my network)
1 Like
Looks like old web app that couldn’t update itself was to blame! Think you’re working now @sandeater
1 Like
This is an awesome tool so far!
1 Like
Hi! I want to congratulate you on the wonderful work you’ve been doing. So far, I’m constantly researching and communicating with the platform to describe and analyze the workouts.
I just need to know what the monthly or annual cost will be! If it’s an affordable price, I’ll leave XERT.
1 Like
I am not able to pull and analyze details for a virtual ride activity. I get the message “cannot find the activity” and suggest to search by ID. When I search by ID it thinks there should be an “i” prefix but I do not have one. Then searching by name and date, it finds the ID but cannot load the detail even though it is public in Strava and exist fully in intervals.icu.
It did work originally but recently I can no longer pull any virtual ride it seems like.
2 Likes
@Bigblob Great project!. Question, what would be the value added of your tool vs DIY via Claude - MCP? … (aside from easier onboarding)
1 Like
Thanks all! Really really appreciate the bug reports, observations, and positive comments.
Pricing - soon! Trying to work it out to be fair and risk free whilst ensuring we can continue improvements - easier said than done! Will have some idea by Friday.
Bugs - noted! On it!
Claude with MCP - we’re doing a lot in the mcp layer of this to make the data more friendly for models, we’re also caching data and doing some trickery there to make things faster and a bit easier on intervalsicu. We could expose it as an mcp, after all it is one under the hood. The real juice comes from our model routing, model selection, model configuration and system prompts at each routed level. This gives speed + cost saving for simpler tasks, and then a much deeper analysis for bigger tasks. I also spent ages tuning model parameters. You could probably get a similar result with Claude - I originally built as chatgpt app but their App Store isn’t ready yet. I also had a way better experience with our system than in that context (I could only test it for myself there - someone posted a pre-made gpt further up this thread you could try!). It was also flipping slow!
If people want an mcp I can try to make that happen, but for now the whole package seems less friction for all users (myself included and I subscribe to all the providers).
3 Likes