85 lines
2.4 KiB
YAML
85 lines
2.4 KiB
YAML
id: https://nde.nl/ontology/hc/class/ThinkingMode
|
|
name: thinking_mode_class
|
|
title: Thinking Mode Class
|
|
prefixes:
|
|
linkml: https://w3id.org/linkml/
|
|
hc: https://nde.nl/ontology/hc/
|
|
schema: http://schema.org/
|
|
prov: http://www.w3.org/ns/prov#
|
|
default_prefix: hc
|
|
imports:
|
|
- linkml:types
|
|
- ../enums/ThinkingModeEnum
|
|
- ../slots/has_description
|
|
- ../slots/has_label
|
|
- ../slots/has_score
|
|
classes:
|
|
ThinkingMode:
|
|
class_uri: schema:PropertyValue
|
|
description: 'LLM thinking mode configuration for API requests.
|
|
|
|
|
|
**DEFINITION**:
|
|
|
|
|
|
ThinkingMode represents the operational mode for LLM reasoning during inference.
|
|
|
|
This class wraps ThinkingModeEnum to provide extensibility and structured metadata.
|
|
|
|
|
|
**GLM 4.7 Thinking Modes** (https://docs.z.ai/guides/capabilities/thinking-mode):
|
|
|
|
|
|
1. **enabled** - Thinking enabled, model reasons before responding
|
|
|
|
2. **disabled** - Thinking disabled, faster responses without reasoning_content
|
|
|
|
3. **interleaved** - Think between tool calls (default since GLM-4.5)
|
|
|
|
4. **preserved** - Retain reasoning across turns (Coding Plan default)
|
|
|
|
|
|
**Ontological Alignment**:
|
|
|
|
- Maps to `schema:PropertyValue` for configuration settings
|
|
|
|
- Enables future extension with timing, tokens, confidence metadata
|
|
|
|
|
|
**Migration Note**:
|
|
|
|
Created 2026-01-16 to replace bespoke `thinking_mode` slot per Rule 53/56.
|
|
|
|
The enum ThinkingModeEnum is preserved; this class adds structure.
|
|
|
|
'
|
|
exact_mappings:
|
|
- schema:PropertyValue
|
|
close_mappings:
|
|
- prov:Plan
|
|
slots:
|
|
- has_label
|
|
- has_description
|
|
- has_score
|
|
slot_usage:
|
|
has_label:
|
|
examples:
|
|
- value: Preserved Thinking
|
|
has_description:
|
|
examples:
|
|
- value: Retains reasoning_content from previous assistant turns in context
|
|
for improved performance.
|
|
annotations:
|
|
specificity_score: 0.75
|
|
specificity_rationale: Highly specific to LLM API configuration domain
|
|
custodian_types: '["D"]'
|
|
custodian_types_rationale: Applies to Digital Platform custodians using LLM
|
|
APIs
|
|
examples:
|
|
- value:
|
|
has_label: Preserved Thinking
|
|
has_description: Retains reasoning across conversation turns for improved
|
|
accuracy.
|
|
- value:
|
|
has_label: Interleaved Thinking
|
|
has_description: Default mode - thinks between tool calls.
|