How It Works
Learn how Push.ai processes and analyzes unstructured data to enhance your Business Graph™.
The Knowledge system in Push.ai processes unstructured data through a pipeline that transforms raw content into a repository of context built around the structured model of your business.
Knowledge is currently only available on the Enterprise Plan.
Artifacts
Artifacts are one of the foundational elements of the knowledge system, representing a piece of unstructured content, like a document, meeting, or Slack channel. Learn more about artifacts and their categories in our Artifacts documentation.
Chunks
Chunks are the atomic building blocks of knowledge processing. We use artifact-specific chunking techniques to break down artifacts into meaningful segments while preserving their original context and structure. This approach enables:
- Semantic Search: Breaking content into appropriately sized chunks allows for more precise and relevant search results
- Context Preservation: Each chunk maintains its relationship to the original artifact and its position within the larger document
- Citation Support: The chunking process preserves metadata and structural information needed for accurate citations
Vector Embeddings and Inference
Once content is chunked, we use vector embeddings to transform text into numerical representations that capture semantic meaning. This enables:
- Semantic Search: Finding relevant content based on meaning rather than just keywords
- Context Mapping: Automatically connecting unstructured content to relevant business objects
- Pattern Recognition: Identifying relationships and patterns across different pieces of content
Business Graph Integration
The processed knowledge is then integrated with your Business Graph™ through:
- Automatic Tagging: Content is automatically tagged with relevant business objects
- Context Enrichment: Unstructured insights are used to enhance the context of structured data
- Relationship Mapping: New relationships between business objects are discovered and mapped