L2: Data Processing
Transform raw validated inputs into structured, enriched data optimized for agent consumption.
What is Layer 2?
Layer 2 (L2) transforms raw validated inputs into structured data that agents can efficiently process. This includes parsing complex formats, normalizing data structures, enriching with contextual information, and aggregating related inputs for batch processing.
Processing Capabilities
Data Parsing
Extract structured data from PDFs, HTML, spreadsheets, and unstructured text. Automatic format detection and conversion.
Normalization
Standardize data formats, clean whitespace, normalize encodings, and apply consistent formatting rules across all inputs.
Aggregation
Group related inputs for batch processing. Combine data from multiple sources into unified payloads for efficient agent execution.
Context Enrichment
Automatically inject relevant context to enhance agent understanding:
- Knowledge Items - Retrieve relevant documents from your Knowledge Base
- Brand Kit Guidelines - Inject voice, tone, and style preferences
- Historical Context - Include past interactions and decisions for continuity
- Variable Substitution - Replace placeholders with dynamic values
Batch Processing
Process multiple inputs in parallel with configurable batch sizes. L2 optimizes payload sizes, manages memory efficiently, and prepares data for downstream layers to maximize throughput without sacrificing quality.