VLM Classifier (AI Chain) icon

VLM Classifier (AI Chain)

Classify document complexity using Vision-Language Models (LOW/HIGH) - For AI Agent chains

VLM Classifier (AI Chain)

AI-generated

Summary

Classifies documents as LOW or HIGH complexity through the configured VLM server and returns the results to an AI chain. Choose whether to attach classification metadata, filter documents by complexity, or do both.

Inputs

  • Model (required) — Select a classifier model listed by the configured Ollama or custom API server.
  • Classification Strategy — Choose Add Metadata Only (default), Filter Documents, or Filter + Metadata.
  • Filter Complexity — When filtering, keep LOW documents, HIGH documents, or both; defaults to both.
  • Request Timeout (ms) — Maximum time for each API request; defaults to 30,000 ms.
  • HIGH Trigger Threshold — P(HIGH) threshold from 0 to 1; defaults to 0.5. Lower values make HIGH classifications easier to trigger. Failed classifications are left unchanged.

Output shape

A classifier provider for an AI chain that accepts documents and returns a list of classified documents.

Documents may provide text through pageContent, text, content, or document, and may include image or base64Image data. Results preserve document fields and add _originalIndex. Metadata strategies also add _complexityClass, _complexityConfidence, _complexityProbabilityHigh, _processingTime, and _modelUsed; failures add _classificationError. Filtering can return an empty list. Classification failures fall back to LOW with zero confidence unless filtered out.

Links

Discussion