<span id="pdf-压缩"></span>

# PDF compression

Compress Catalog PDFs before parsing and pass the returned file sources to downstream nodes. Use a sample to check readability, especially small text, stamps, and table rules.

<span id="配置节点"></span>

## Configure nodes

Choose **PDF Compress** under AI Processing. Select Catalog files directly or bind the source output of [Read MOI Volume](../data-io/read-volume.md).

| Setting | Description |
| --- | --- |
| File | Select one or more Catalog files; upstream sources can bind to `sources`. |
| Quality `quality` | `high`, `medium`, or `low`; default: `high`. Lower levels reduce image clarity, so check small text, stamps, and table borders. |
| Target size `target_bytes` | Size in bytes. Default: `0`, meaning no target is specified. This is a processing target, not a guaranteed result size. |

Use one source input method rather than combining independent selections. The node receives file references, not parsed text.

<span id="连接后续解析"></span>

## Continue with parsing

```text
读取 MOI 卷 → PDF 压缩 → 文档解析 → 分段
```

1. Select a scanned PDF and bind the read node's `sources` to the PDF compression source input.
2. Keep high resolution selected initially, and leave the target size at `0`.
Start with high quality and no target. Bind the compression node's returned `sources` to [Document Parsing](document-parse.md). Keeping the original read node bound downstream would continue parsing the original input.
4. After execution, check the compression record, file size, and preview, then check the parsed text, tables, and page numbers.

Non-PDF files and PDFs that cannot be made smaller pass through unchanged. Successful execution does not mean every file was compressed. Image re-encoding can affect OCR, so validate a representative sample before expanding the run.

<span id="检查结果"></span>

## Check results

| Output | Purpose |
| --- | --- |
| `sources` | File sources used for subsequent parsing. |
| `files` | Per-file compression ledger for checking original/returned files, original/processed sizes, compression strategies, and target achievement. |
| `summary` | Summary of this processing operation. |

Non-PDF files and files that cannot shrink pass through unchanged. Node success does not mean every file was compressed. If a target size is set, also inspect `target_met` in the ledger. Without a target, this value does not indicate compression failure.

Compression may re-encode page images. To retain details for OCR or extraction, verify a representative file before processing the rest.
