Jev model capability selection

Match a requested output to a model capability

Compare the requested output with supplied fictional text, judgment and image profiles. Select a compatible profile or identify a missing capability; selecting a profile does not invoke it.

Match the output requirement to a stated ability.Live API
Try an example

Original synthetic input. Edit the task or evidence to make a new API request.

Source: Workflow inspiration on X

Source and example scope

The inputs, rubric and preview are original. This window demonstrates the stated decision only; it does not reproduce the source application or its measurements.

Which supplied fictional profile has the required output capability?

Decision rules and labels
Text writerTyped judgeImage rendererNo capability matchNeed requirements
A real response from the Jev API

Edit any input to explore a different decision. No request is sent until you run it.

Your next decision

Not run yet

Choose an example or write your own input, then run it to see the ranked choices.

Results are not stored by this site. Your submitted text is processed by the API provider.

The decision

How does Jev model capability selection use the supplied evidence?

Jev model capability selection compares an output requirement with three fictional profiles. A request for release-note paragraphs needs prose generation; a request for one supplied department label needs a typed judgment. The model selects a profile ID and the local preview displays it. This exercise never invokes a second model.

An API-collected X post discussed selecting between text generation and typed decisions. Our original catalog isolates that capability distinction without importing the post's platform versions, package support or model performance claims. Text writer, Typed judge and Image renderer are invented profiles with only the abilities explicitly shown in their JSON.

A profile name is a poor substitute for a contract. The fictional judge can select a label but cannot produce a paragraph. The writer produces prose, while the renderer produces an image. The decision receives those profiles beside the task so the request can be inspected without guessing what a real provider offers.

Original Jev model capability selection workflow diagram showing supplied evidence, a typed decision and the local preview.
Original workflow illustration. The output is a local preview, not an executed action or source screenshot.
How it works

Try the example in four steps

Use the sample to understand the rule, then test your own input for model capability matching.

  1. Illustration of adding the text or evidence for a decision

    Add the information to check

    Compare the requested output with supplied fictional text, judgment and image profiles.

  2. Illustration of defining a question and its allowed choices

    Confirm the question and labels

    Open the decision rules and keep the intended distinction between Text writer, Typed judge, Image renderer, No capability match, Need requirements. Edit the choices only when testing your own rule.

  3. Illustration of sending one input and comparing the returned choices

    Run the decision

    Use a sample first, then try your own input. Compare the selected label with the other returned probabilities.

  4. Illustration of reviewing a result and an uncertain item before using it

    Review the next step

    Review Need requirements before using a result. The label is a decision about the supplied information, not an executed action.

Our API test

What the six original inputs returned.

What the six original inputs returned.
Original test inputObserved labelLeading probability
Generate release-note textText writer99%
Pick a typed labelTyped judge98%
Generate a visualImage renderer98%
Absent execution capabilityNo capability match68%
Missing output requirementNeed requirements91%
Bounded classificationTyped judge79%
  • Measured adapter round trips: 466-1176 ms. Estimated input cost: $0.00002638-$0.00002671 per request. These observations are not a latency guarantee or invoice.

6 actual Jev API calls on October 1, 2026. Original synthetic inputs check this rubric and handoff; they do not establish held-out accuracy. Probabilities are model outputs.Verified

Understand the results

The release-note task asks for actual paragraphs

The release-note task asks for actual paragraphs.

Read the example notes

Its expected label is Text writer even though a separate judge could assess whether the eventual paragraphs satisfy a rubric. Choosing the generator and judging its result are different calls. The selected profile contains prose-generation; no release note is written by this demo.

Pick a typed label

The department task supplies a fixed list and explicitly asks for a label without a response paragraph.

Read the example notes

Typed judge is the supplied capability that matches this output. The second classification sample checks the same boundary with another wording. These checks concern the requested output type, not the accuracy of a downstream department classifier.

The image task requires a raster output, so its profile is Image renderer

The image task requires a raster output, so its profile is Image renderer.

Read the example notes

A text model could help draft a prompt, but that would not fulfill the stated requirement to produce the image. Our window sends only a Choice request and returns a local profile preview. It neither creates an image nor uploads a media asset.

Source and example scope

The deployment task requests infrastructure execution that none of these profiles offers.

Read the example notes

No capability match retains that absence instead of selecting the most plausible model name. The task could become supportable after a separately authorized executor is added, but the catalog and question would then be different and require a fresh evaluation.

Using the result in your application

The vague task gives no output requirement.

Read the example notes

Need requirements is separate from a known request whose capability is absent. An application can ask whether the user wants prose, an image or a fixed decision before selection. The offline result remains pending and contains no inferred best model or fabricated profile.

Using the result in your application

A production selector needs verified provider capabilities, availability, authorization, payload validation and measured costs.

Read the example notes

The semantic label here checks only the supplied ability. The v2 consumer binds the selected fixed payload to the exact candidate JSON in that request. Altering a profile ID or capability invalidates an old observation; replay does not contact that profile.

Using the result in your application

Download the exact sample requests and saved observations below.

Read the example notes

Expected labels are original fixture checks and stay outside the request sent to Jev. The observation file preserves the complete probabilities and the hash of each exact request. Editing a task, rubric or candidate requires a fresh response; a matching sample name does not make an older answer current.

Using the result in your application

The JavaScript runner makes one independent SDK request for a chosen sample ID.

Read the example notes

It reads TYPESAFE_API_KEY privately from the environment and has no automatic retry. The Python consumer operates offline: it checks the rubric, labels, unique sample identities, request hashes and valid probability distributions before selecting a fixed local preview. A tied leading result remains pending for review.

Source and example scope

The page reports measured adapter round trips and estimated input cost for the recorded calls.

Read the example notes

Those values describe these requests, not a provider latency guarantee or a billing statement. Six deliberately authored fixtures can check the stated behavior, but they do not establish accuracy on an unseen dataset. Keep disagreements and missing-input examples when testing an integration.

Runnable offline example

Replay one request and inspect the saved preview.

View the code example
Replay commands
npm install @typesafe-ai/sdk@0.6.0
# Set TYPESAFE_API_KEY privately in your shell.
node run-choice-sample.mjs samples.json prose
python preview_choice_workflow_v2.py samples.json saved-api-results.json
Scope of this demo

Where does this decision fit?

Useful with explicit supplied evidence

  • Select a supplied fictional model profile by the output capability a task requires: prose generation, typed text judgment or image generation. Preserve absent capabilities and missing requirements. Original capability catalog only; downstream models are not called and no actual Apple model claim is adopted.
  • Inspecting the original task and a fixed output before integrating external tools.
  • Comparing an explicit expected label with the complete recorded distribution.

Needs separate implementation or evidence

  • Inferring facts absent from the supplied input.
  • Treating a semantic selection as authorization to act.
  • Executing an external action or generalizing six fixtures into an accuracy claim.
Example FAQ

Questions about this original decision workflow.

Keep relevance, compliance and execution separate.

Are these real deployed model profiles?

No. They are an original fictional capability catalog. The page makes no claim about a named provider, operating system or available deployment.

Does the selected profile run after the decision?

No. The preview prints its supplied metadata. Downstream authentication, transport and generation are separate application work.

Can a typed judgment write my release notes?

The fictional Typed judge profile selects a supplied label only. The release-note task needs the separate prose-generation capability.

Why keep unsupported and unspecified requests separate?

A deployment request has a known capability missing from the catalog. An unspecified task needs an output requirement before a match can be assessed.

Do these results measure the best model?

No. They test a stated capability contract on six original inputs. They do not compare provider accuracy, latency, availability or price.

Try a task from your own workflow

Change the supplied task or evidence here, then inspect the live response. Playground supports another explicit decision.