POST/api/v1/evals/safe-regenerate

Eval -> regenerate loop with cost-budget enforcement

Evaluates the input on per-dimension rubrics; if the worst dim is below the threshold, asks the BYOK LLM to regenerate, then re-evaluates. Stops on pass, max iterations, or no-improvement. Pre-flight cost estimate gates against `maxCostUsd` (402 if exceeded). Records final outcome to `safe_regenerate_runs` + `cost_ledger`.

Authentication

Send Authorization: Bearer YOUR_API_KEY on every request. Generate API keys at /dashboard/api-keys.

Request body required

Example

{
  "content": "<The candidate output to evaluate.>",
  "prompt": "<The prompt that produced the content.>",
  "threshold": 0.8,
  "maxIterations": 3,
  "dimensions": [
    "string"
  ],
  "scorerSet": "basic",
  "regeneratorModel": "openai/gpt-4o-mini",
  "maxCostUsd": 0.5,
  "projectId": "00000000-0000-0000-0000-000000000000",
  "recordHistory": false
}
Schema
{
  "application/json": {
    "schema": {
      "type": "object",
      "required": [
        "content",
        "prompt"
      ],
      "properties": {
        "content": {
          "type": "string",
          "description": "The candidate output to evaluate."
        },
        "prompt": {
          "type": "string",
          "description": "The prompt that produced the content."
        },
        "threshold": {
          "type": "number",
          "minimum": 0,
          "maximum": 1,
          "default": 0.8
        },
        "maxIterations": {
          "type": "integer",
          "minimum": 1,
          "maximum": 10,
          "default": 3
        },
        "dimensions": {
          "type": "array",
          "items": {
            "type": "string"
          },
          "description": "Subset of pillars to gate on (default: all in scorerSet)."
        },
        "scorerSet": {
          "type": "string",
          "enum": [
            "basic",
            "deep"
          ],
          "default": "basic"
        },
        "regeneratorModel": {
          "type": "string",
          "default": "openai/gpt-4o-mini"
        },
        "maxCostUsd": {
          "type": "number",
          "minimum": 0,
          "maximum": 100,
          "default": 0.5
        },
        "projectId": {
          "type": "string",
          "format": "uuid"
        },
        "recordHistory": {
          "type": "boolean",
          "default": false
        }
      }
    }
  }
}

Response

200 example

{
  "result": {
    "content": "string",
    "finalScore": 0,
    "passedThreshold": false,
    "iterationsTaken": 0,
    "iterationScores": [
      0
    ],
    "stopReason": "passed"
  },
  "cost": {
    "estimatedWorstCaseUsd": 0,
    "actualUsd": 0,
    "llmCallCount": 0,
    "actualInputTokens": 0,
    "actualOutputTokens": 0
  }
}

All status codes

200Evaluation + regen result.
400(no description)
401(no description)
402Payment required — estimated cost exceeds the per-call budget cap.
429(no description)

Code samples

cURL

curl -X POST \
  https://evalguard.ai/api/v1/evals/safe-regenerate \
  -H "Authorization: Bearer $EVALGUARD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "content": "<The candidate output to evaluate.>", "prompt": "<The prompt that produced the content.>", "threshold": 0.8, "maxIterations": 3, "dimensions": [ "string" ], "scorerSet": "basic", "regeneratorModel": "openai/gpt-4o-mini", "maxCostUsd": 0.5, "projectId": "00000000-0000-0000-0000-000000000000", "recordHistory": false }'

TypeScript

import { EvalGuard } from "@evalguard/sdk";

const client = new EvalGuard({ apiKey: process.env.EVALGUARD_API_KEY });

const response = await client.request({
  method: "POST",
  path: "/api/v1/evals/safe-regenerate",
  body: {
    "content": "<The candidate output to evaluate.>",
    "prompt": "<The prompt that produced the content.>",
    "threshold": 0.8,
    "maxIterations": 3,
    "dimensions": [
      "string"
    ],
    "scorerSet": "basic",
    "regeneratorModel": "openai/gpt-4o-mini",
    "maxCostUsd": 0.5,
    "projectId": "00000000-0000-0000-0000-000000000000",
    "recordHistory": false
  },
});
console.log(response);

Python

from evalguard import EvalGuard
import os

client = EvalGuard(api_key=os.environ["EVALGUARD_API_KEY"])

response = client.request(
    method="POST",
    path="/api/v1/evals/safe-regenerate",
    body={
    "content": "<The candidate output to evaluate.>",
    "prompt": "<The prompt that produced the content.>",
    "threshold": 0.8,
    "maxIterations": 3,
    "dimensions": [
        "string"
    ],
    "scorerSet": "basic",
    "regeneratorModel": "openai/gpt-4o-mini",
    "maxCostUsd": 0.5,
    "projectId": "00000000-0000-0000-0000-000000000000",
    "recordHistory": False
},
)
print(response)

Go

package main

import (
	"context"
	"fmt"
	"os"

	"github.com/evalguard/evalguard-go"
)

func main() {
	client := evalguard.NewClient(os.Getenv("EVALGUARD_API_KEY"))
	resp, err := client.Request(context.Background(), "POST", "/api/v1/evals/safe-regenerate", map[string]any{"content": "<The candidate output to evaluate.>", "prompt": "<The prompt that produced the content.>", "threshold": 0.8, "maxIterations": 3, "dimensions": []any{"string"}, "scorerSet": "basic", "regeneratorModel": "openai/gpt-4o-mini", "maxCostUsd": 0.5, "projectId": "00000000-0000-0000-0000-000000000000", "recordHistory": false})
	if err != nil { panic(err) }
	fmt.Println(resp)
}

Errors

400401402429

Other Evaluations endpoints