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

# Advanced search

# Advanced Search Failovers & PDF Document Parsers

Hiroshi OS supports multi-engine search fallbacks (SearXNG with DuckDuckGo fallback) and local/remote binary PDF text matrix extraction.

## ReAct XML Interfaces

### Multi-Engine Search

Execute queries across localized SearXNG instances or fallback DuckDuckGo HTML indices:

```xml theme={null}
<search engine="searxng">rust async web server</search>
```

### PDF Text Layout Extraction

Perform memory-mapped structural layout parses on workspace PDF documents:

```xml theme={null}
<pdf_extract path="docs/references/api_manual.pdf"/>
```

## Configurations

Configure the default timeout thresholds, SearXNG endpoint URLs, and safety size constraints inside your settings file:

```yaml theme={null}
advanced_search:
  searxng_url: "http://localhost:8080"
  fallback_to_ddg: true
  max_pdf_size_bytes: 15728640 # 15MB standard protection cap
```

## Performance Metrics

| Measurement Parameter Axis      | Boundary Limits                                                           |
| ------------------------------- | ------------------------------------------------------------------------- |
| **PDF Matrix Extraction Speed** | Parsing and restructuring a dense 5MB layout file executes in **\< 15ms** |
| **Search Deduplication Index**  | Merging and cleaning alternate cluster engines completes in **\< 8ms**    |