Skip to content

simodocs

temprl.org Console

HTTP API

Response format

JSON

{
  "id": "evt_8sKq2...",
  "model": "simo-1-pro",
  "answers": {
    "refund": {"type": "noul", "noul": 0.991},
    "team":   {"type": "choice", "choice": "billing",
               "probabilities": {"billing": 0.962, "shipping": 0.021, "tech": 0.017}, "confidence": 0.962},
    "mood":   {"type": "score", "score": 2.31,
               "legend": {"0": "calm", "1": "annoyed", "2": "angry", "3": "furious"},
               "probabilities": {"0": 0.02, "1": 0.11, "2": 0.41, "3": 0.46}, "confidence": 0.46},
    "amount": {"type": "fill", "value": "89", "parsed": 89.0, "valid": true, "confidence": 0.99}
  },
  "usage": {"input_tokens": 212, "output_tokens": 0, "cost": 0.00000636}
}

Field	Description
`id`	The request id, also sent as `x-request-id`. Quote it to support
`model`	The model that answered
`answers`	One answer per question id, plus one per `derive` name
`usage.input_tokens`	Input tokens read (what you pay for)
`usage.output_tokens`	Output tokens written (free)
`usage.cost`	Cost of this request in US dollars

Other fields (more usage detail, and service metadata under simo) may appear. Ignore what you don't use.

Answers by type

`type`	Fields
`noul`	`noul`: P(true), in [0, 1]
`choice`	`choice`: the most likely option. `probabilities`: `{option: p}`, summing to 1. `confidence`. `args`: `{name: answer}` for the chosen option, when the question had `args`
`score`	`score`: the expected level, 0-based. `legend`: `{"0": level, ...}`. `probabilities`: `{"0": p, ...}`, one per level. `confidence`: the largest of them
`fill`	`value`: the text. `parsed`: the typed value (int, number or list) when `kind` says so. `valid`. `confidence`. `check` with `self_check`
`extract`	`value` (or `values` with `many`). `valid`. `confidence`. `check` with `self_check`
extract, `kind: "coordinates"`	`coordinates`: `[x1, y1, x2, y2]` or `null`. `center`: `[x, y]` or `null`. `image_size`: `[width, height]`. `valid`. `confidence`

Decision probabilities (noul, probabilities) are calibrated. confidence on reads is not; check is.

Coordinates are in the pixels of the image as sent. If you resized or padded an image before sending it, map the answer back yourself (the Python SDK does this for you).

Headers

Header	Example	Description
`x-request-id`	`evt_8sKq2...`	The request id
`x-simo-cost-usd`	`0.000006360`	What this request cost
`x-simo-balance-usd`	`24.999993640`	Your balance after it
`x-simo-latency-ms`	`412`	Time spent in the API

These headers are exposed to browsers through CORS, though keys belong on your server.