Function createVisionPipeline

Create a vision pipeline with sensible defaults by auto-detecting which providers are installed in the current environment.

The factory probes for optional peer dependencies (ppu-paddle-ocr, tesseract.js, @huggingface/transformers) and cloud API keys in environment variables, then configures the pipeline accordingly.

Auto-detection logic

Check	Result
ppu-paddle-ocr importable	ocr: 'paddle'
tesseract.js importable (paddle missing)	ocr: 'tesseract'
Neither importable	ocr: 'none'
@huggingface/transformers importable	handwriting: true, documentAI: true, embedding: true
OPENAI_API_KEY / ANTHROPIC_API_KEY / etc. set	cloudProvider configured

Caller-supplied config overrides always take precedence over auto-detected values.

Example

// Full auto-detection
const pipeline = await createVisionPipeline();

// Override strategy but auto-detect everything else
const localOnly = await createVisionPipeline({
  strategy: 'local-only',
});

// Explicit full config (no auto-detection)
const explicit = await createVisionPipeline({
  strategy: 'progressive',
  ocr: 'tesseract',
  handwriting: true,
  documentAI: false,
  embedding: true,
  cloudProvider: 'openai',
  cloudModel: 'gpt-4o',
  confidenceThreshold: 0.85,
});
Copy