openai/codex · error · ConstraintError

To use model `{model}`, you need to use auto review.

Error message

To use model `{model}`, you need to use auto review.

What it means

ConstraintError::AutoReviewRequired (codex-rs/config/src/constraint.rs:19) fires when a requirements layer restricts a model so it may only be used together with auto review (the automatic approvals-reviewer mode). Selecting that model without auto review enabled fails config assembly rather than silently downgrading the safety requirement.

Source

Thrown at codex-rs/config/src/constraint.rs:19

use std::fmt;
use std::sync::Arc;

use crate::config_requirements::RequirementSource;
use thiserror::Error;

#[derive(Debug, Error, PartialEq, Eq)]
pub enum ConstraintError {
    #[error(
        "invalid value for `{field_name}`: `{candidate}` is not in the allowed set {allowed} (set by {requirement_source})"
    )]
    InvalidValue {
        field_name: &'static str,
        candidate: String,
        allowed: String,
        requirement_source: RequirementSource,
    },

    #[error("To use model `{model}`, you need to use auto review.")]
    AutoReviewRequired { model: String },

    #[error("field `{field_name}` cannot be empty")]
    EmptyField { field_name: String },

    #[error("invalid rules in requirements (set by {requirement_source}): {reason}")]
    ExecPolicyParse {
        requirement_source: RequirementSource,
        reason: String,
    },

    #[error(
        "invalid requirement for MCP server `{server_name}` (set by {requirement_source}): {reason}"
    )]
    McpServerRequirementParse {
        server_name: String,
        requirement_source: RequirementSource,
        reason: String,

View on GitHub (pinned to 339751715c)

Solutions

1. Enable auto review for the run (set the approvals reviewer to auto) and retry
2. Or switch to a model that is not gated by the auto-review requirement
3. Or ask the requirements owner to lift the auto-review constraint for that model

Example fix

# before
model = "gpt-5.1-codex-max"
# (auto review not configured)

# after
model = "gpt-5.1-codex-max"
approvals_reviewer = "auto"

// If requirements are introspectable, gate the model pick up front:
if model_requires_auto_review(&requirements, &model)? && !auto_review_enabled {
    return Err("enable auto review or pick another model".into());
}

match assemble_config(...) {
    Err(ConstraintError::AutoReviewRequired { model }) => {
        prompt_user(format!("{model} requires auto review; enable it or pick another model"));
    }
    r => r?,
}

When it happens

Trigger: Setting model via config.toml, the -m flag, or an API override to a model whose requirement enforces auto review, while the approvals reviewer is not set to auto; the model constraint validator returns AutoReviewRequired during config load or set.

Common situations: Orgs gate high-autonomy models behind automatic code review; users pin such a model in a dotfiles-shared config that does not enable auto review; profiles copied between machines where the reviewer setting differs.

Related errors

• invalid value for `{field_name}`: `{candidate}` is not in th
• field `{field_name}` cannot be empty
• failed to decrypt encrypted task id
• `bypass_hook_trust` override must be a boolean
• remote control URL cannot be a base

AI-assisted analysis of openai/codex@339751715c (2026-08-25). Data as JSON: /api/errors/db81256098f1e3bf. Report an issue: GitHub.