25 lines
1001 B
Plaintext
25 lines
1001 B
Plaintext
|
## NAPI-RS Implementation Details
|
||
|
|
|
||
|
|
### Native Performance
|
||
|
|
|
||
|
|
This binding uses NAPI-RS to provide native Node.js bindings with:
|
||
|
|
|
||
|
|
- **Zero-copy data transfer** between JavaScript and Rust layers
|
||
|
|
- **Native thread pool** for concurrent document processing
|
||
|
|
- **Direct memory management** for efficient large document handling
|
||
|
|
- **Binary-compatible** pre-built native modules across platforms
|
||
|
|
|
||
|
|
### Threading Model
|
||
|
|
|
||
|
|
- Single documents are processed synchronously or asynchronously in a dedicated thread
|
||
|
|
- Batch operations distribute work across available CPU cores
|
||
|
|
- Thread count is configurable but defaults to system CPU count
|
||
|
|
- Long-running extractions block the event loop unless using async APIs
|
||
|
|
|
||
|
|
### Memory Management
|
||
|
|
|
||
|
|
- Large documents (> 100 MB) are streamed to avoid loading entirely into memory
|
||
|
|
- Temporary files are created in system temp directory for extraction
|
||
|
|
- Memory is automatically released after extraction completion
|
||
|
|
- ONNX models are cached in memory for repeated embeddings operations
|