agent-loopsComing Soon Spec

Browser-Use DOM Element Targeter

Choice-based next-element targeting for browser agents with a blocking-overlay Noul.

Architecture overview

Web agents click many times per task. This page documents using Jev AI tools to pick a target element from a candidate list and detect blocking overlays — keeping each step typed and cheap on output tokens.

Traditional LLM limitation

Sending large HTML blobs to generative LLMs on every click is slow and expensive.

Jev AI tools advantage

Pass summarized candidates in state; Choice returns one of your element IDs. Pricing follows official TypeSafe rates.

Official API schema (field names from docs.typesafe.ai)

Source: TypeSafe HTTP API reference. Question IDs below this table are directory example compositions — you choose them.

FieldTypeRangeNotes
statestring | object | arrayPlain text or structured JSONContent to evaluate. Shared across all questions in one request.
modelstringe.g. "jev-latest"Optional; defaults to TypeSafe flagship alias when omitted in SDKs.
questions.<id>.type"choice" | "score" | "noul"Exactly one of three primitivesQuestion ID is chosen by you; answers return under the same keys.
questions.<id>.instructionsstringNatural-language judgmentThe actual question sent for inference (IDs are not sent to the model).
questions.<id>.criteria (choice)Record<option, string | null>1–255 optionsMap of option key → rubric description.
questions.<id>.criteria (score)string[]≥ 2 ordered levelsOrdered level descriptions; score is a weighted position along them.
questions.<id>.criteria (noul){ true?: string; false?: string }OptionalOptional clarification of yes/no meanings. Answer field is noul ∈ [0, 1].
answers.<id> (choice){ type, choice, probabilities, confidence }choice ∈ criteria keys; confidence ∈ [0, 1]Probabilities sum to 1 across options.
answers.<id> (score){ type, score, legend, probabilities, confidence }score may fall between levelslegend maps level index → description.
answers.<id> (noul){ type, noul }noul ∈ [0, 1]Probability that the answer is yes. No separate confidence field.
usage{ input_tokens, output_tokens }Non-negative integersToken accounting for the request.

Example composition for this Jev AI tools workflow

Sample state and question keys are illustrative compositions for this directory page — not a separate official product API. Primitives remain Choice / Score / Noul.

DOM Snapshot: Checkout modal with candidates [Place Order #btn-1], [Quantity +], [Continue Shopping]. Goal: Complete purchase.
Choice
target_element: Next interactive element target
Options: [place_order, quantity_stepper, continue_shopping, close_modal]
Noul
is_blocked: Is a blocking overlay preventing standard interaction?

Runnable call examples

Endpoint: https://api.typesafe.ai/v1/systemone (official). Requires your own TYPESAFE_API_KEY.

curl

curl -X POST https://api.typesafe.ai/v1/systemone \
  -H "Authorization: Bearer $TYPESAFE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"state":{"goal":"Complete purchase","candidates":[{"id":"place_order","label":"Place Order"}]},"model":"jev-latest","questions":{"target":{"type":"choice","instructions":"Next interactive element","criteria":{"place_order":"Place Order","quantity_stepper":"Quantity +","continue_shopping":"Continue","close_modal":"Close"}},"is_blocked":{"type":"noul","instructions":"Blocking overlay visible?"}}}'

TypeScript

import { choice, noul, TypeSafeClient } from "@typesafe-ai/sdk";

const client = new TypeSafeClient();
const step = await client.systemOne({
  state: { goal, candidates },
  model: "jev-latest",
  questions: {
    target: choice("Next interactive element", Object.fromEntries(
      candidates.map((c) => [c.id, c.label])
    )),
    is_blocked: noul("Blocking overlay visible?"),
  },
});

Python

from typesafe_sdk import Choice, Noul, TypeSafeClient

with TypeSafeClient() as client:
    step = client.system_one(
        state={"goal": goal, "candidates": candidates},
        model="jev-latest",
        questions={
            "target": Choice(
                instructions="Next interactive element",
                criteria={c["id"]: c["label"] for c in candidates},
            ),
            "is_blocked": Noul(instructions="Blocking overlay visible?"),
        },
    )

Latency & cost (source attribution)

Official end-to-end latency range: ~70–500ms; many calls land near ~100ms from US West Coast (source: typesafe.ai / TypeSafe public materials, 2026-09). Official list price: $0.042 / 1M input tokens; output tokens free (source: typesafe.ai, as of 2026-09). Card latency figures are illustrative compositions within that published range — not independent lab measurements by this directory.

  • Card illustration on this page: ~84ms (illustrative, within official range — not a lab run by jevaitools.com).
  • Official list price: $0.042 / 1M input tokens; output tokens free (source: typesafe.ai, as of 2026-09).

This directory has not published an independent measurement script for this page. To measure yourself: call the official endpoint with your key, record wall-clock p50/p95 andusage.input_tokens, and keep the date of the run.

Comparison with generative LLMs on the same decision task

TypeSafe publishes workflow evaluations where Jev is compared with frontier LLMs on accuracy, cost, and latency (company materials, 2026). Those multipliers are vendor-reported ceilings, not results measured by this directory. Run the same questions through an LLM structured-output adapter on your labeled set before choosing a stack.

Suitable for

  • Closed candidate sets extracted by a DOM parser
  • Hot loops where output-token cost must stay near zero
  • Overlay detection before click attempts

Not suitable for

  • Pixel-level vision without a candidate list
  • Writing browser automation scripts
  • CAPTCHA solving

Common failure modes

  • Stale candidates after a SPA re-render
  • Dumping full HTML instead of labeled candidates
  • No other/none option when the goal element is missing

Other Jev AI Tools

This site is an independent third-party directory and is not affiliated with, endorsed by, or operated by TypeSafe AI.