# What is a ShinrAI record?

One document is rounded up separately in blocks of 1,000 Unicode code points (minimum one block for accepted empty text). Standard uses 1 record per block, Batch 0.5 and Real-Time 1.6. Spaces and punctuation count. Code points are not UTF-8 bytes or model tokens; a visual character can contain several code points.

| Submitted text | Standard | Batch | Real-Time |
| --- | ---: | ---: | ---: |
| 1,000 code points | 1 | 0.5 | 1.6 |
| 1,001 code points | 2 | 1 | 3.2 |
| 2,500 code points | 3 | 1.5 | 4.8 |
| Two separate 500-code-point documents | 2 | 1 | 3.2 |

A chatbot submitting a 1,000-character question plus four 1,500-character passages as one text uses seven Standard records each time. 8,000 questions use 56,000 records. Batch-indexing 1,000 new files of 3,000 characters adds 1,500 records: 57,500 total. Initial library indexing is a separate workload. Text submitted again, including conversation history, uses records again. Restoring values locally does not itself consume another API record.

## Editable planning examples

The estimator starts with a few everyday quantities. Open Advanced assumptions for text length, context, AI steps and processing tier. Employee examples use a fixed 20 working days per month. All figures below are illustrative Standard workloads; file indexing uses Batch.

| Example | Assumptions | Records per month | Matching monthly plan |
| --- | --- | ---: | --- |
| Customer workflows | 10,000 tickets/documents × 2 AI steps × 3,000 characters | 60,000 | Team |
| Everyday AI use | 50 employees × 20 working days × 10 messages/day × 4,000 characters including context | 40,000 | Starter |
| Company knowledge chatbot | 50 employees × 20 days × 8 questions/day; question plus retrieved context is 7,000 characters; 1,000 files at 3,000 characters indexed in Batch | 57,500 | Team |
| Online shop AI assistant | 100,000 customer conversations × 3 AI steps × 6,000 characters including order/conversation context | 1,800,000 | Business |
| Large app customer support | 500,000 customer conversations × 3 AI steps × 4,000 characters including account/conversation context | 6,000,000 | Platform |

For everyday AI use, increasing the team to 100 employees doubles the example to 80,000 records and recommends Team. Recommendations use the smallest allowance that covers the selected workload and processing tier. Changing the assumptions changes the result.

The shop scenario covers order questions, return guidance and reply preparation. The app scenario covers account/service questions, response preparation and handover to human support. Integration with a particular shop or app is scoped separately. These sample volumes are planning inputs, not measured ShinrAI customer usage.

Each subscription month begins on its billing date. Monthly allowance is used before packs and does not roll over. Purchased managed packs do not expire. Offline packs are separate, installation-bound usage. API responses expose usage.weighted_records; Dashboard shows remaining allowance. Request-size and rate limits are separate from the billing record definition.

Interactive examples: https://shinrai.innovius.io/records
Estimator: https://shinrai.innovius.io/#estimate
API documentation: https://shinrai.innovius.io/docs