Skip to main content

​
Parameters

ParameterTypeDefaultDescription
agentAgentRequiredPre-configured Agent class for cognitive processing
configOptional[AgenticChunkingConfig]NoneConfiguration object with all settings

​
Functions

​
__init__

Initialize agentic chunker. Parameters:
  • agent (Agent): Pre-configured Agent class for cognitive processing
  • config (Optional[AgenticChunkingConfig]): Configuration object with all settings

​
_chunk_document

Synchronous chunking - delegates to async implementation. Parameters:
  • document (Document): Document to chunk
Returns:
  • List[Chunk]: List of chunks

​
_achunk_document

Core agentic chunking pipeline. Parameters:
  • document (Document): Document to process with AI agents
Returns:
  • List[Chunk]: List of cognitively-optimized chunks with rich metadata

​
_get_cache_key

Generate cache key for text content. Parameters:
  • text (str): Text content to generate key for
Returns:
  • str: MD5 hash of the text content

​
_extract_propositions

Extract atomic propositions from document using AI agent. Parameters:
  • document (Document): Document to extract propositions from
Returns:
  • List[str]: List of extracted propositions

​
_validate_propositions

Validate proposition quality and filter invalid ones. Parameters:
  • propositions (List[str]): List of propositions to validate
Returns:
  • List[str]: List of validated propositions

​
_group_propositions_into_topics

Group propositions into coherent thematic topics using AI agent. Parameters:
  • propositions (List[str]): List of propositions to group
  • document (Document): Source document
Returns:
  • List[Topic]: List of topics containing grouped propositions

​
_validate_topic_sizes

Validate topic sizes and split oversized topics. Parameters:
  • topics (List[Topic]): List of topics to validate
Returns:
  • List[Topic]: List of validated topics

​
_optimize_topics

Optimize topic assignments by merging small topics. Parameters:
  • topics (List[Topic]): List of topics to optimize
Returns:
  • List[Topic]: List of optimized topics

​
_find_chunk_indices_in_document

Find start and end indices of chunk text within document. Parameters:
  • chunk_text (str): Text content of the chunk
  • document_content (str): Full document content
Returns:
  • Tuple[int, int]: Start and end indices

​
_create_chunks_from_topics

Create chunks from grouped topics. Parameters:
  • topics (List[Topic]): List of topics to create chunks from
  • document (Document): Source document
Returns:
  • List[Chunk]: List of created chunks

​
_finalize_topic_chunk

Finalize a chunk from topics with metadata enrichment. Parameters:
  • topics (List[Topic]): List of topics for the chunk
  • document (Document): Source document
Returns:
  • Chunk: Finalized chunk with enriched metadata

​
_refine_topic_metadata

Refine topic metadata using AI agent. Parameters:
  • chunk_text (str): Text content of the chunk
  • topic (Topic): Base topic for metadata
Returns:
  • RefinedTopic: Refined topic with title and summary

​
_score_chunk_coherence

Score chunk coherence and add quality assessment. Parameters:
  • chunks (List[Chunk]): List of chunks to score
Returns:
  • List[Chunk]: List of chunks with coherence scores

​
_calculate_coherence_score

Calculate coherence score for a chunk. Parameters:
  • chunk (Chunk): Chunk to calculate score for
Returns:
  • float: Coherence score between 0 and 1

​
_assess_chunk_quality

Assess chunk quality based on coherence score. Parameters:
  • coherence_score (float): Coherence score to assess
Returns:
  • str: Quality assessment (“excellent”, “good”, “fair”, “poor”)

​
_fallback_chunking

Fallback to recursive chunking on agent failure. Parameters:
  • document (Document): Document to chunk with fallback method
Returns:
  • List[Chunk]: List of chunks created with fallback method

​
_add_processing_metadata

Add processing metadata to chunks. Parameters:
  • chunks (List[Chunk]): List of chunks to add metadata to
  • processing_time (float): Processing time in milliseconds

​
get_agentic_stats

Get statistics about agentic processing. Returns:
  • Dict[str, any]: Dictionary containing processing statistics

​
clear_agentic_caches

Clear all agentic processing caches.
⌘I