Decision models

POST https://api.routerplex.com/v1/systemone

Decision models answer typed questions about a piece of content instead of generating text. Send a state and a map of named questions; get back one typed answer per question, with probabilities and confidence your code can branch on. Every question is evaluated in parallel and in isolation against the same state, so adding questions barely changes latency.

Use the same API key and base URL as for chat models. Decision models are not available through /v1/chat/completions.

Models #

Launch promotion: decision models are free for a limited time. The struck-through rate is the list price RouterPlex will charge when the promotion ends. Output tokens are always free. Calls still need an account with an active balance.

jev-latest is TypeSafe's current stable release and jev-preview its newest build; both point to jev-1.13.0 today. The response's model field names the version that answered. If you tune confidence thresholds against a version, pin jev-1.13.0.

Request #

curl https://api.routerplex.com/v1/systemone \
  -H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "jev-latest",
    "state": "Help! My payouts have been failing for 3 days.",
    "questions": {
      "is_urgent": {"type": "noul", "instructions": "Does this convey urgency?"},
      "team": {
        "type": "choice",
        "instructions": "Which team should handle this?",
        "criteria": {"billing": "Payments, refunds", "technical": "Bugs, outages", "sales": "Pricing"}
      },
      "frustration": {
        "type": "score",
        "instructions": "How frustrated is the customer?",
        "criteria": ["Calm", "Frustrated", "Very angry"]
      }
    }
  }'

Question types #

instructions and criteria also accept objects, so a question can carry the data it refers to: {"question": "Is this the same person?", "candidate": {...}}.

Response #

{
  "model": "jev-1.13.0",
  "answers": {
    "is_urgent": {"type": "noul", "noul": 0.95},
    "team": {
      "type": "choice", "choice": "billing", "confidence": 0.96,
      "probabilities": {"billing": 0.98, "technical": 0.02, "sales": 0.0}
    },
    "frustration": {
      "type": "score", "score": 1.04, "confidence": 0.95,
      "legend": {"0": "Calm", "1": "Frustrated", "2": "Very angry"},
      "probabilities": {"0": 0.0, "1": 0.96, "2": 0.04}
    }
  },
  "usage": {"input_tokens": 382, "output_tokens": 73, "cost": 0.0}
}

usage.cost is the amount billed in USD, also sent as the x-routerplex-cost header. Every response carries x-routerplex-request-id, which you can look up with the cost & usage API.

Designing questions #

Ask atomic questions: each one should be a judgment a knowledgeable person could make in a few seconds. Break a broad judgment ("rate this pitch") into separate questions (market size, feasibility, differentiation) and combine the answers in code, where you control the weights. Use confidence as a second axis: act automatically on confident answers, and route uncertain ones to a person or a larger model.

Using the TypeSafe SDK #

The request and response shapes match TypeSafe's System One API, so TypeSafe's Python SDK (tested with typesafe-sdk 0.7.2) works with RouterPlex once you set its base URL to https://api.routerplex.com (no /v1; the SDK adds it) and pass your RouterPlex key. Use base_url= or the TYPESAFE_BASE_URL environment variable. The JavaScript SDK takes the same root as baseURL.

import os
from typesafe_sdk import Choice, Noul, TypeSafeClient
 
client = TypeSafeClient(
    base_url="https://api.routerplex.com",
    api_key=os.environ["ROUTERPLEX_API_KEY"],
)
result = client.system_one(
    model="jev-latest",
    state="I was charged twice this month and nobody answers my emails.",
    questions={
        "billing": Noul(instructions="Is this a billing problem?"),
        "team": Choice(instructions="Which team?", criteria={"billing": "Payments", "technical": "Bugs"}),
    },
)
print(result.answers["team"].choice, result.answers["team"].confidence)

Errors #