wally decisions

wally decisions scores questions with Eve, the decision model. It sends one request to POST /v1/decisions and prints a probability for every answer. wally decide is an alias for the same command.

wally decisions (alias wally decide) sends every request to Eve (eve), RunAnywhere’s decision model, by default. You don’t need to pass a model.

It runs on your Wally sign-in (wally account login). There is no key to set.

Example

wally decisions \
  --input 'After the 2.3 deploy, every order confirmation email shows the wrong currency for EU customers.' \
  --ask 'Is this a bug report?' \
  --choice 'Which team owns this?=frontend,payments,email' \
  --score 'How severe is it?=cosmetic,minor,major,outage'

Output from a production run on 2026-10-04 (wally 0.7.3):

Is this a bug report?
yes  99.0%
  yes          ████████████████████  99.0%
  no                                  1.0%

Which team owns this?
email  56.4%
  frontend                            2.3%
  payments     ████████              41.3%
  email        ███████████           56.4%

How severe is it?
major  70.9%
  cosmetic     █                      6.8%
  minor        ███                   16.8%
  major        ██████████████        70.9%
  outage       █                      5.5%

Each question shows its most probable answer, then a bar for every label in the order you gave them. One status line goes to stderr, with the tokens billed, the cost and the round-trip time from your machine:

eve · <tokens> tokens · $<cost> · <ms>ms

Flags

Flag	Short	Argument	Default	What it does
`--model`	`-m`	model id	`eve`	The decision model to ask. Eve unless you pass another; you never need to set it for Eve
`--input`	`-i`	text		The text every question is about
`--input-file`		path, or `-`		Read the input from a file, or from stdin with `-`
`--ask`		`'QUESTION'`		A yes/no question. Repeatable.
`--choice`		`'QUESTION=a,b,c'`		A choice question with 2 to 26 options. Repeatable.
`--score`		`'QUESTION=l0,l1,…'`		A score question with 2 to 10 levels, labelled 0 to 9 in the order given. Repeatable.
`--request`		path, or `-`		Send a complete request body from a file, or from stdin with `-`. See Request files.
`--temperature`		number	1	Probability temperature, greater than 0 and at most 100
`--json`			off	Print the raw response JSON instead of the bars
`--no-retry`			off	Do not retry HTTP 429 or 503
`--help`	`-h`			Show help

Every flag in the code is in --help; there are no hidden ones. Wally’s global --json works here too.

Input. Give exactly one of --input, --input-file or piped stdin. --input and --input-file together are refused. With neither, Wally reads stdin if it is not a terminal.
Questions. Mix --ask, --choice and --score freely, 1 to 128 in all. They are sent in the order given, with ids q1, q2, and so on.
Retries. Without --no-retry, a 429 or 503 is retried, honouring Retry-After up to 10 seconds, and an expired sign-in is refreshed once.

Request files

--request takes the exact body of POST /v1/decisions, with your own question ids. It is the only way to set prompt_format_version, an option description, or the yes and no text of a yes/no question.

{
  "model": "eve",
  "input": "After the 2.3 deploy, every order confirmation email shows the wrong currency for EU customers.",
  "questions": [
    { "id": "is_bug", "type": "yes_no", "question": "Is this a bug report?" },
    {
      "id": "team",
      "type": "choice",
      "question": "Which team owns this?",
      "options": [{ "name": "frontend" }, { "name": "payments" }, { "name": "email" }]
    },
    {
      "id": "severity",
      "type": "score",
      "question": "How severe is it?",
      "levels": ["cosmetic", "minor", "major", "outage"]
    }
  ]
}

wally decisions --json --request request.json

Wally checks the file against the contract before sending it, so a misspelled field fails locally instead of being dropped. --request cannot be combined with --input, --input-file, --ask, --choice, --score, --temperature or --model; put those in the file.

Exit codes

Code	Meaning
`0`	Answered
`1`	Refused by the server, or a runtime error: not signed in, `403 model_not_entitled`, a question over the token window, an unserved `prompt_format_version`
`2`	A usage error caught before any request: blank input, no question, an unknown field in a request file, more than 128 questions, more than 26 options

Eve is not a chat model

Chat commands refuse it before anything is sent:

$ wally opencode -m eve
Error: eve is a decision model, not a chat model; use `wally decisions -m eve` instead

Get started

Going further

Example

Flags

Request files

Exit codes

Eve is not a chat model

​Example

​Flags

​Request files

​Exit codes

​Eve is not a chat model

Example

Flags

Request files

Exit codes

Eve is not a chat model