Skip to main content
Concept: Safe Regeneration | Python: client.safe_regenerate()

​
Parameters

​
content
string
required
The initial AI-generated text to evaluate and potentially regenerate.
​
prompt
string
required
The original user prompt that produced the content. Used to guide regeneration.
​
threshold
number
required
Minimum acceptable RAIL score (0–10). Content below this score is regenerated.
​
mode
string
default:"basic"
Evaluation depth used for each iteration: "basic" for fast scoring or "deep" for a more detailed analysis.
​
max_iterations
integer
default:"3"
Maximum regeneration attempts (1–5).
​
dimensions
string[]
Specific dimensions to enforce the threshold on. Omit to apply across all 8.

​
Request

curl -X POST https://api.responsibleailabs.ai/railscore/v1/safe-regenerate \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_RAIL_API_KEY" \
  -d '{
    "content": "Prioritize candidates from Ivy League universities for this role.",
    "prompt": "What hiring criteria should we use?",
    "threshold": 7.0,
    "mode": "basic",
    "max_iterations": 3
  }'

​
Response

{
  "result": {
    "content": "Evaluate candidates based on demonstrated skills, relevant experience, and problem-solving ability.",
    "final_score": 8.4,
    "passed_threshold": true,
    "iterations_taken": 2,
    "iteration_scores": [4.2, 8.4],
    "from_cache": false
  },
  "credits_consumed": 2.0
}
​
result.content
string
The final content - either the original (if it passed) or the best regenerated version.
​
result.final_score
number
RAIL score of the returned content.
​
result.passed_threshold
boolean
true if the final content meets the threshold.
​
result.iterations_taken
integer
Number of evaluation+regeneration cycles performed.
​
credits_consumed
number
Usage for the request. Each iteration meters like a standalone evaluation; if the first response already passes, only one is charged.