Documentation Index
Fetch the complete documentation index at: https://developer.kodexa.ai/llms.txt
Use this file to discover all available pages before exploring further.
A Knowledge Item Type is a template that defines a configurable capability in the system. While Knowledge Feature Types describe what you know about documents, Knowledge Item Types describe what you can do - such as customizing extraction prompts, applying validation rules, or modifying processing behavior.
When Do You Need Item Types?
Create a Knowledge Item Type when you want to:
- Allow customization of extraction prompts for specific data elements
- Define validation rules that can be selectively applied
- Create processing behaviors that can be configured per document type
- Enable business users to customize system behavior without code changes
How Item Types Work
- Define the Type - Create a Knowledge Item Type that describes a capability (e.g., “Prompt Override”)
- Create Items - Create configured instances with specific values (e.g., “Use this prompt for invoice totals”)
- Connect via Sets - Knowledge Sets link Features to Items, triggering behaviors based on document characteristics
Item Type Structure
| Field | Purpose | Example |
|---|
slug | Unique identifier | prompt-override |
name | Display name | Prompt Override |
description | What this capability does | Customize the extraction prompt for a data element |
options | Configuration parameters users can set | targetField, promptText, model |
Options Define Configuration
Options specify what parameters users can configure when creating items:
options:
- name: targetField
type: string
label: Target Field
description: The data element this prompt applies to
required: true
- name: promptText
type: text
label: Prompt Text
description: The custom prompt to use for extraction
required: true
- name: temperature
type: number
label: Temperature
description: LLM temperature setting (0-1)
default: 0.1
Creating an Item Type
Via YAML (GitOps)
Create a file in kodexa-resources/knowledge-item-types/:
# kodexa-resources/knowledge-item-types/prompt-override.yaml
slug: prompt-override
name: Prompt Override
description: Customize the extraction prompt for a specific data element
options:
- name: targetField
type: string
label: Target Field
description: The taxon path this prompt applies to
required: true
- name: promptText
type: text
label: Prompt Text
description: The custom prompt to use for extraction
required: true
- name: includeExamples
type: boolean
label: Include Examples
description: Whether to include few-shot examples
default: true
# manifests/main.yaml
resources:
knowledge-item-types:
- prompt-override
Via API
curl -X POST "https://platform.kodexa.ai/api/knowledgeItemTypes" \
-H "Authorization: Bearer $KODEXA_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"slug": "prompt-override",
"name": "Prompt Override",
"description": "Customize the extraction prompt for a specific data element",
"options": [
{
"name": "targetField",
"type": "string",
"label": "Target Field",
"required": true
},
{
"name": "promptText",
"type": "text",
"label": "Prompt Text",
"required": true
}
]
}'
Via Kodexa UI
- Navigate to Knowledge in the sidebar
- Click Item Types
- Click Create Item Type
- Fill in name, slug, description
- Add options to define configurable parameters
- Save
Creating Knowledge Items
Once you have an Item Type, create configured items:
Via YAML
# Knowledge Item configuration
title: SEC 10K Extraction Prompt
description: Specialized prompt for extracting data from SEC 10K filings
knowledgeItemType: prompt-override
active: true
properties:
targetField: "financial/revenue"
promptText: |
Extract the total revenue figure from this SEC 10K filing.
Look for "Total Revenue", "Net Revenue", or "Total Sales" in the
financial statements section. Return the most recent fiscal year value.
includeExamples: true
Via API
curl -X POST "https://platform.kodexa.ai/api/knowledgeItems" \
-H "Authorization: Bearer $KODEXA_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"title": "SEC 10K Extraction Prompt",
"description": "Specialized prompt for extracting data from SEC 10K filings",
"knowledgeItemTypeSlug": "prompt-override",
"active": true,
"properties": {
"targetField": "financial/revenue",
"promptText": "Extract the total revenue figure from this SEC 10K filing...",
"includeExamples": true
}
}'
Connecting Items to Features via Knowledge Sets
Knowledge Items don’t do anything on their own - they need to be connected to documents through Knowledge Sets. A Knowledge Set says “when a document has these features, apply these items.”
See Knowledge Sets for details on creating sets.
Common Item Type Examples
Validation Rule
slug: validation-rule
name: Validation Rule
description: Apply custom validation logic to extracted data
options:
- name: targetField
type: string
label: Target Field
required: true
- name: ruleType
type: select
label: Rule Type
options:
- value: required
label: Required Field
- value: range
label: Numeric Range
- value: pattern
label: Regex Pattern
- value: custom
label: Custom Expression
- name: ruleExpression
type: string
label: Rule Expression
description: The validation expression or pattern
- name: errorMessage
type: string
label: Error Message
description: Message shown when validation fails
Field Ignore Rule
slug: field-ignore
name: Field Ignore Rule
description: Skip extraction for specific fields under certain conditions
options:
- name: targetField
type: string
label: Target Field
required: true
- name: reason
type: string
label: Reason
description: Why this field should be ignored
Processing Priority
slug: processing-priority
name: Processing Priority
description: Adjust processing priority for specific document types
options:
- name: priority
type: select
label: Priority Level
options:
- value: high
label: High Priority
- value: normal
label: Normal
- value: low
label: Low Priority
- name: slaHours
type: number
label: SLA (Hours)
description: Target processing time in hours
Item Types vs Feature Types
| Aspect | Feature Type | Item Type |
|---|
| Purpose | Capture metadata about documents | Define configurable behaviors |
| Direction | Information FROM documents | Actions applied TO documents |
| Examples | Vendor, Document Type, Language | Prompt Override, Validation Rule |
| Persistence | Features are reused across documents | Items are applied via Knowledge Sets |
| Created by | Often by agents during processing | Typically by developers/admins |
Next Steps
