NeuroForge
Appearance

Use your device setting or keep a preference on this browser.

Contact
Navigate
Support the labTalk to Lloyd

Quick find

Find what you need

Esc

Try a job or topic, such as meeting notes.

    Search learning resources Free lessons, printables and activities.

    ERAIS / Reasoning Lab

    Work through a decision
    you can check.

    Choose an example, change its facts and test the questions provided. Inspect the results, including when information is missing.

    Availability not checkedRead the recorded controlled-English proof

    The legacy route uses its existing browser verification.

    Choose an example
    Text and mixed examples

    These examples use JSON and need a qualified semantic provider. Check Service details for live availability.

    Change the request details, then check which rules apply.

    Give the system a rule and a few facts.

    Facts included

    Unchecked means unknown, not false.

    Maximum 512 bytes

    Ctrl / Cmd + Enter to run. Drafts are kept in this tab across reloads. Reset example replaces its draft. Nothing runs until you ask.

    Answers

    Not run

    Change a fact or keep the example as it is, then check the answers.

    These checks use your supplied facts and rules. They do not verify whether those facts are true in the real world.

    Service details and request timing

    Formal: unchecked · Semantic: unchecked

    Browser round trip
    Not measured
    Origin round trip
    Not reported
    Decision HTTP calls
    0
    Upstream calls
    Not reported
    Questions in batch
    0
    Answered / abstained
    0 / 0

    Calibration: unchecked. Formal 0/1 distributions are logical indicators, not learned probabilities or certainty about real-world truth. Semantic label-token scores are uncalibrated and are not formal proofs. Type-valid does not mean correct.

    No request receipt yet.

    Use the public API

    From experiment to code

    Use the decision API in your app.

    Check service availability, then use the curl request to call the API. It updates with valid edits. No key, cookie, account or browser challenge is required.

    Request JSON Schema

    64 questions, 16 options per question, 64 KiB input. Host-wide capacity can return busy. Edge and origin deadlines: 2 seconds. No unbounded queue or promise of unlimited compute.

    Timings include network legs, not just inference. Health checks are excluded from decision-call counts. A measured request is not a benchmark.

    curl / application/json
    curl --fail-with-body 'https://neuroforge.io/api/rapid-decision/v1/decide' \
  -H 'Content-Type: application/json' \
  --data-binary @- <<'ERAIS_REQUEST'
{
  "contract": "erais.rapid-decision.request.v1",
  "state": {
    "variables": {
      "service_down": "1",
      "waiting_hours": "30",
      "repeat_contacts": "3"
    }
  },
  "questions": {
    "escalate": {
      "type": "noul",
      "predicate": {
        "left": "waiting_hours",
        "op": "ge",
        "right": "24"
      }
    }
  }
}
ERAIS_REQUEST

    Static one-question example. Check service availability before use.

    Capabilities and limits

    Use exact comparisons (eq/ne/lt/le/gt/ge), atom queries and all/any/not. Truth uses "noul"; Score is a zero-based level, not a continuous rating. Choice/Score criteria must be known and match uniquely. Otherwise values are null: read status before acting.

    No matched Jev comparison is presented. Formal deductions do not establish semantic intelligence or calibrated confidence.

    Try controlled English with a proof trace · Reasoning Lab examples. Public prototype. Check important answers.

    Continue exploring