Attributes

Overview

The Attributes page gives you a unified view of every data point ContractRabbit has extracted across your document portfolio. Each row represents a single extracted value tied to a specific document.

Two layers of extraction

ContractRabbit extracts attributes at two levels of granularity:

Document-level attributes

When a file is processed, an LLM call extracts headline attributes from the full document text: parties, effective date, expiration date, governing law, contract value, duration, and other fields defined by attribute prompts specific to the document type. If an effective date and duration are found but no expiration date, ContractRabbit computes the expiration deterministically.

Clause-level attributes

After document-level extraction, every clause in the document is processed in parallel batches. Each clause is analyzed for specific typed values — a dollar amount mentioned in a liability cap, a date in a renewal clause, a duration in a non-compete restriction. This gives you granular, per-provision data that document-level extraction alone would miss.

Attribute types

ContractRabbit extracts the following types of attributes, each with structured metadata:

Each attribute type is color-coded for easy scanning.

Smart extraction rules

• Entity exclusion — Party names from the document are provided to the LLM as an exclusion list, preventing aliases like "the Company" or "Employer" from being misclassified as entity attributes
• Corporate suffix detection — Only proper names with corporate suffixes (Inc, LLC, Corp, Ltd, GmbH, plc) or full personal names are classified as entities
• Computed attributes — When an effective date and duration are found but no expiration date exists, the expiration is calculated and stored automatically

Table columns

• File — Link to the source document
• Document Title — Title of the document
• Document Type — Classification badge (e.g., NDA)
• Attribute Type — Color-coded type indicator
• Canonical Path — Where in the document structure the attribute was found (e.g., K-NonCompete.Duration)
• Extracted Value — The actual value, with type-specific metadata (e.g., currency for Money)
• Context — Surrounding text (hover to see full context)
• Created — When the attribute was extracted

Filtering

Click the filter button to open the filter panel. Available filters include:

General filters

• Text search — Full-text search across all attributes
• File — Filter to a specific document
• Attribute type — Select one or more types to display

Type-specific filters

When you select a single attribute type, additional filters appear:

• Duration — Range slider with min/max days
• Money — Range slider with currency selection
• Date — Date range picker
• Numeric — Range slider
• Entity — Text search with entity kind checkboxes
• Citation / Location / Class — Text search

Canonical path filter

Type a path prefix to filter by document structure location. Suggestions appear as you type. Canonical paths follow a hierarchical naming convention (e.g., K-Indemnification.MonetaryCap), so filtering by a prefix returns all attributes within that category.

Active filters appear as badges above the table. Click the x on any badge to remove it, or use Clear all to reset.

Using attributes with the Agent

The Agent can query attributes directly using SQL-level aggregations:

• "What is the average non-compete duration across our NDAs?"
• "Show me the 95th percentile contract value by document type"
• "Build a table of all service agreements with duration, governing law, and contract value"

The Agent discovers available canonical paths dynamically, so you don't need to know the exact path names — just describe what you're looking for in natural language.

Pagination

Choose from 25, 50, or 100 rows per page using the dropdown at the bottom of the table.