Adaptive’s zero completion insurance automatically protects you from being charged for failed requests or responses with no output tokens. When responses are empty or encounter errors, you pay nothing - even if underlying providers charge for processing.

​
How It Works

1

Request Monitoring

Every response is automatically analyzed for completion tokens and status
2

Failure Detection

Identifies responses with zero output tokens and blank or error finish reasons
3

Automatic Protection

Credits are not deducted from your account when protection criteria are met
4

Cost Absorption

Adaptive absorbs provider costs for protected requests at no charge to you

​
Protection Criteria

Zero completion insurance automatically applies when responses meet either of these conditions:

Zero Output + Blank Finish

Response has zero completion tokens AND finish reason is blank/nullCommon scenarios: Provider timeouts, processing errors, system failures

Error Finish Reason

Response has an error finish reason regardless of token countCommon scenarios: Content policy violations, rate limits, service unavailable

​
Real Examples

Request: “Write a detailed analysis of quantum computing”
Response: Empty (provider timeout)
Finish Reason: null
Your Cost: **0.00∗∗(protected)∗∗WithoutInsurance∗∗:0.00** (protected) **Without Insurance**: 0.12 for prompt processing

​
Automatic Coverage

Always Enabled

No configuration required - protection is automatically active for all accounts

All Models

Universal coverage across all providers and model types without exceptions

Zero Cost

Completely free - no additional fees or charges for insurance coverage

​
Coverage Examples

​
Protected Scenarios

These response types are automatically covered at no cost: 
{
  "id": "chatcmpl-abc123",
  "choices": [{
    "message": {"content": ""},
    "finish_reason": null
  }],
  "usage": {
    "prompt_tokens": 15,
    "completion_tokens": 0,  // Zero output tokens
    "total_tokens": 15
  }
}
// Cost: $0.00 (protected)

​
Normal Billing

These responses are billed normally as they contain successful completions: 
{
  "id": "chatcmpl-ghi789",
  "choices": [{
    "message": {"content": "Hello! How can I help you today?"},
    "finish_reason": "stop"  // Normal completion
  }],
  "usage": {
    "prompt_tokens": 9,
    "completion_tokens": 12,  // Has output tokens
    "total_tokens": 21
  }
}
// Cost: Normal billing applies

​
Viewing Protected Requests

​
Dashboard Activity

In your Adaptive dashboard, protected requests are clearly marked:

Zero Cost Display

Credit Deduction: Shows $0.00 for protected requests

Protection Badge

Visual Indicator: Protected requests display insurance badge

Reason Details

Protection Reason: Shows why request was protected (error type, empty response)

Provider Costs

Transparency: View actual provider charges that were absorbed

​
Response Metadata

Protected requests include additional metadata in the response: 
{
  "id": "chatcmpl-protected",
  "choices": [{"message": {"content": ""}}],
  "usage": {"prompt_tokens": 12, "completion_tokens": 0},
  "insurance": {
    "protected": true,           // Request was protected
    "reason": "zero_completion", // Why it was protected
    "original_cost": 0.08,       // What you would have paid
    "savings": 0.08              // Amount saved
  }
}

​
Insurance Benefits

Risk-Free Testing

Experiment freely without fear of charges for failed requests

Cost Predictability

Budget control - only pay for successful completions

Provider Reliability

Protection from outages - no charges during service disruptions

Development Safety

Safe debugging - test requests without unexpected costs

​
Insurance Statistics

Coverage Rate

5-15% of requests typically qualify for protection

Average Savings

$0.05-0.20 per protected request depending on prompt complexity

Detection Speed

Instant protection applied automatically without delays

​
Comparison with Other Providers

Most LLM providers charge for prompt processing regardless of response success. Adaptive is unique in providing comprehensive zero completion insurance.
ProviderFailed Request ChargesEmpty Response ChargesError State Charges
Adaptive$0.00$0.00$0.00
OpenAIPrompt tokens chargedPrompt tokens chargedPrompt tokens charged
AnthropicPrompt tokens chargedPrompt tokens chargedPrompt tokens charged
Google AIPrompt tokens chargedPrompt tokens chargedPrompt tokens charged
Other APIsPrompt tokens chargedPrompt tokens chargedPrompt tokens charged

​
FAQ

​
Best Practices

Test Freely

Experiment confidently knowing failed requests won’t generate unexpected charges

Monitor Usage

Track protection rates in your dashboard to understand request patterns

Handle Errors Gracefully

Implement proper error handling for protected requests in your application

Use for Development

Safe development environment - test edge cases without cost concerns

​
Implementation

Zero completion insurance requires no code changes - it works automatically with your existing integrations: 
// Normal request - insurance automatically active
const completion = await openai.chat.completions.create({
  model: "gpt-4",
  messages: [{ role: "user", content: "Hello!" }]
});

// Check if request was protected
if (completion.insurance?.protected) {
  console.log(`Saved $${completion.insurance.savings} on failed request`);
}

​
Next Steps

Cost Monitoring

View your insurance savings and protected request statistics

Provider Resiliency

Learn about automatic failover and reliability features