technicalbillingother
01 / ChoicePick a label.
Choose from the options you define.
Free access to TypeSafe Jev Messy words.
Messy words.
Clear decisions.
Is this urgent? Which label fits? How positive is it?
Give your app useful answers, not another paragraph.
Free for learning and testing. No account needed to explore.
1Add some text2Ask one thing3Get a useful answer
The playground
A small decision. Try one.
No setup. Start with an example.
Pick a label
Sample mode shows an illustrative answer. No API call.
3Answer
IllustrativeThe best-fitting label
refund
refund94%
delivery4%
other2%
Model confidence0.88
Probabilities compare your labels. Confidence describes how decisive the model is, not whether it is correct.
Samples stay in your browser. Live API sends your input to typesafe.pro and TypeSafe. Please do not paste private information.
How it works
Three small primitives.
Lots of useful decisions.
yesno
02 / NoulCheck a yes / no.
Get the probability that the answer is yes.
03 / Score
Put it on a scale.
Define levels, then get a score between them.
Jev makes a judgement. Your code decides what happens next. Meet the primitives
Examples
Small problems
worth automating.
Choice
Sort support messages
Send a customer to the right kind of help, without reading every message by hand.
customer messagerefund
Noul
Spot an urgent message
Notice time-sensitive messages so they can get a closer look sooner.
a messageyes probability
Score
Understand a review
Put feedback on a scale, from unhappy to delighted, using levels you define.
product reviewa score
Choice
Organise study notes
Give each note a subject label, so the right revision material is easier to find.
a study notebiology
Noul
Find refund requests
Check one clear thing: is the customer actually asking for their money back?
customer textyes probability
Choice
Choose an agent tool
Match a request to an available tool. Keep the actual tool call in your code.
user requestcalculator
Showing 6 of 9 examples
From "oh, I get it" to your app
Now put it
in your app.
The same example you just tried.
In a language you already use.
YOUR API BASE URL
https://api.typesafe.proAnonymous HTTP examples; no key required by this gateway.
One request. Named, typed answers.
Change the playground. The code follows.
Your playground, in code
import jsonfrom urllib.error import HTTPError, URLErrorfrom urllib.request import Request, urlopenBASE_URL = "https://api.typesafe.pro"payload = { "state": "Hi! I ordered the blue trainers, but they don't fit. Could I get my money back?", "model": "jev-latest", "questions": { "request_type": { "type": "choice", "instructions": "What is the customer asking about?", "criteria": { "refund": "The customer wants money returned for an order.", "delivery": "The customer asks about shipping, tracking, or when an order will arrive.", "other": "A question that is not about a refund or delivery.", }, }, },}request = Request( f"{BASE_URL}/v1/systemone", data=json.dumps(payload).encode("utf-8"), headers={"Content-Type": "application/json"}, method="POST",)try: with urlopen(request, timeout=20) as response: result = json.load(response)except HTTPError as error: raise SystemExit(f"API returned HTTP {error.code}") from errorexcept URLError as error: raise SystemExit(f"Connection failed: {error.reason}") from errorprint(json.dumps(result["answers"], indent=2))Already using a TypeSafe SDK? Change the API root.
The official SDKs require a key. Use your own typesafe.pro gateway token, stored server-side in TYPESAFE_PRO_API_KEY. For keyless requests, use the HTTP examples above. Do not use an upstream provider key.
Good questions
A little more
context.
No mystery. Here is what you are trying, and where it fits.
No. Jev answers small, focused questions about text. It returns a label, a score, or a probability, not a written reply. Use a generative model for writing, and use these answers for decisions in your code.
You can explore the examples without either. typesafe.pro is designed for free learning and testing, including anonymous HTTP access. Live access is subject to gateway availability and fair-use limits.
Yes: switch the playground to Live API. Sample mode contains hand-written examples to explain the interface; it does not run a model or evaluate new text.
Yes. A high probability or confidence value is not proof that a decision is correct. Try representative examples, keep the original input, and send important or unclear decisions to a person.
You can edit the full JSON request, change the model, define your own labels or score levels, and ask multiple independent questions about the same input.
No. typesafe.pro is an independent access gateway. Jev is developed by TypeSafe AI, and the linked TypeSafe documentation describes their upstream models and API.
Start with one sentence.
See where it takes you.
No setup to get through. Just a small thing to try.
Make your first decision