AI Transaction Detection

Can be one of the following types:

The Chat Parameters class defines a set of options for tuning the behavior of LLM completion requests in Grooper. These parameters allow users to control the randomness, diversity, and topicality of generated responses, enabling tailored outputs for a wide range of scenarios.

Typical usage involves adjusting one or more parameters to guide the model’s output toward the desired style or content. For example, increasing randomness for creative tasks, or reducing repetition for summarization.

Configuration Guidance

• Adjust either 'Temperature' or 'Top P' to control randomness and diversity, but not both simultaneously.
• Use 'Presence Penalty' and 'Frequency Penalty' together to reduce repetition and encourage broader topic coverage.

Usage in Grooper

Chat Parameters are used throughout Grooper to influence LLM completions for document processing, data extraction, and other AI-driven tasks.

Impact

Proper configuration of ChatParameters can significantly affect the quality, creativity, and reliability of LLM outputs. Experimentation is recommended to find the best settings for your specific use case.

Can be one of the following types:

Quoting methods define how content is selected from a document and formatted as input for large language model (LLM) completions. Their primary role is to extract, transform, and present relevant document content—referred to as a "quote"—to the AI in a way that maximizes context, accuracy, and efficiency.

Theory and Purpose

Modern LLMs are highly sensitive to the context and structure of their input. The quality and relevance of the information provided to the model directly impacts its ability to generate accurate, useful responses. Quoting methods serve as the bridge between raw document data and the AI, ensuring that only the most pertinent, well-structured, and cost-effective content is included in the prompt.

• Targeted Context: By selecting only the necessary portions of a document (such as a specific field, table, or region), quoting methods help focus the model's attention and reduce noise.
• Cost Efficiency: LLMs have input token limits and usage costs. Quoting methods help minimize prompt size by including only what is needed, avoiding unnecessary or redundant content.
• Context Awareness: Some quoting methods apply preprocessing (such as tab or paragraph marking) or structure the data (such as JSON or layout schemas) to help the AI better interpret complex layouts, tables, or natural language.
• Adaptability: Different tasks (extraction, summarization, classification, etc.) may require different quoting strategies. The quoting method can be selected or configured to match the needs of the operation.

How Quoting Methods Work

A quoting method typically:

• Defines the scope of content to extract (entire document, specific region, extracted values, etc.).
• Optionally applies preprocessing or formatting to the content.
• Packages the result in a format suitable for the AI (plain text, JSON, layout schema, etc.).
• Supplies the quote as part of the prompt to the LLM, often with a standard prefix or instructions.

Examples

• Extracted: Selects all or part of the document, optionally applying preprocessing.
• Labeled Region: Extracts content following a header label, useful for tables or sections.
• Data Values: Supplies the full Data Model hierarchy as JSON.
• Layout Objects: Provides a structured map of the document's visual layout.