Class ChunkerScaffold

java.lang.Object
ai.pipestream.module.pipelineprobe.moduletest.scaffold.ChunkerScaffold
All Implemented Interfaces:
StepScaffold
public class ChunkerScaffold extends Object implements StepScaffold
Chunker scaffold: the raw chunker input (echo output) carries no vector-set directives, so the chunker no-ops ("nothing to chunk"). In the real pipeline the engine's NodeConfigDirectivePopulator stamps directives onto the doc from the chunker node's custom_config.json_config — the door bypasses the engine, so this scaffold performs the SAME stamp with the SAME config: DefaultChunkerConfig.sentence10x3OnBody(boolean)(true), the exact preset every pipeline-crawl runs (sentence-10-3 on body × minilm + paraphrase-minilm), converted Struct→proto the way the engine converts it.

  • Constructor Details

    • ChunkerScaffold

      public ChunkerScaffold(DefaultChunkerConfig chunkerConfig)
      Creates the chunker scaffold, precomputing the vector-set directives from the sentence10x3OnBody(true) preset so each wrapped doc carries the same directives the engine would stamp.
      Parameters:
      chunkerConfig - the default chunker config used to derive the directives

  • Method Details

    • module

      public String module()
      Description copied from interface: StepScaffold
      The module/step this scaffold targets (e.g. "chunker").
      Specified by:
      module in interface StepScaffold
      Returns:
      the module/step name this scaffold targets

    • wrap

      public ai.pipestream.data.v1.PipeStream wrap(ai.pipestream.data.v1.PipeDoc doc)
      Description copied from interface: StepScaffold
      Turn a corpus doc into the step's expected input PipeStream (directives, context_params, …).
      Specified by:
      wrap in interface StepScaffold
      Parameters:
      doc - the corpus document to wrap
      Returns:
      the PipeStream shaped as the step expects