> ## Documentation Index
> Fetch the complete documentation index at: https://docs.medera.info/llms.txt
> Use this file to discover all available pages before exploring further.

# How it Works

> Request structure and best practices

The Code Prediction endpoint is stateless and accepts either a `documentId` or a free-form `text` block.

```bash theme={"system"}
POST /api/coding/predict
{
  "code_systems": ["icd-10-cm", "cpt"],
  "context": {
    "documentId": "doc_abc"
  }
}

```

Or:

```bash theme={"system"}
POST /api/coding/predict
{
  "code_systems": ["icd-10-cm"],
  "context": {
    "text": "Patient presents with persistent low mood..."
  }
}

```

## Best practices

* Always pass `documentId` over `text` when available — span anchors require the document context

* Restrict `code_systems` to what you need; more systems = more work

* Set `min_confidence` to filter low-confidence predictions

* For CDI, use `/api/coding/cdi/review` not `/predict`