Documentation

Safe Regeneration

Evaluate content and iteratively regenerate until quality thresholds are met. The API evaluates, generates a RAIL-guided prompt, regenerates, and re-evaluates up to the configured maximum iterations.

client.safeRegenerate()

Server-side regeneration

const result = await client.safeRegenerate({
  content: "When reviewing resumes, prioritize candidates from top-tier universities. Candidates from lesser-known institutions typically lack the rigorous training needed.",
  mode: "basic",
  maxRegenerations: 3,
  thresholds: {
    overall: { score: 8.0, confidence: 0.5 }
  },
  domain: "general"
});

console.log(result.status);               // "passed" | "max_iterations_reached"
console.log(result.best_content);         // Improved content
console.log(result.best_iteration);       // Which iteration was best
console.log(result.best_scores.rail_score.score);
console.log(result.credits_consumed);

result.iteration_history?.forEach(iter => {
  console.log(`Iteration ${iter.iteration}: ${iter.scores.rail_score.score} (passed: ${iter.thresholds_met})`);
});

Client-side regeneration

Use your own LLM to regenerate. The API returns a RAIL-guided prompt and session ID; you regenerate then submit the result via safeRegenerateContinue().

// Step 1: Start session — API evaluates and returns a guided prompt
const initial = await client.safeRegenerate({
  content: "Content that needs improvement...",
  maxRegenerations: 3,
  thresholds: { overall: { score: 8.0 } }
});

if (initial.status === "awaiting_regeneration" && initial.rail_prompt) {
  const { system_prompt, user_prompt } = initial.rail_prompt;

  // Regenerate with your model
  const completion = await openai.chat.completions.create({
    model: "gpt-4o",
    messages: [
      { role: "system", content: system_prompt },
      { role: "user",   content: user_prompt }
    ]
  });

  // Step 2: Submit regenerated content for re-evaluation
  const continued = await client.safeRegenerateContinue({
    sessionId: initial.session_id,
    regeneratedContent: completion.choices[0].message.content
  });

  console.log(continued.status);       // "passed" or "awaiting_regeneration"
  console.log(continued.best_content); // Best content so far
}

Sessions expire after 15 minutes. Expired sessions throw a SessionExpiredError.

Parameters

ParameterTypeDefaultDescription
contentstringRequiredText to evaluate and improve (10–10,000 chars)
modestring"basic""basic" or "deep"
maxRegenerationsnumber3Maximum iterations (1–5)
thresholdsobjectoverall ≥ 7.0Threshold config
domainstring"general"Content domain for context-aware scoring

Response: SafeRegenerateResult

{
  status: "passed",              // "passed" | "max_iterations_reached" | "awaiting_regeneration"
  original_content: "...",
  best_content: "...",
  best_iteration: 2,
  best_scores: {
    rail_score: { score: 8.4, confidence: 0.82 },
    dimension_scores: { ... },
    thresholds_met: true
  },
  iteration_history: [
    { iteration: 1, thresholds_met: false, failing_dimensions: ["fairness"] },
    { iteration: 2, thresholds_met: true,  failing_dimensions: [] }
  ],
  credits_consumed: 4.0,
  credits_breakdown: { evaluations: 2.0, regenerations: 2.0, total: 4.0 }
}
← EvaluationCompliance →