Manage domain-specific terminology and synonyms to improve AI answer accuracy
The Glossary feature systematically manages domain-specific terms, abbreviations, and business rules used within your organization.
It enables AI to accurately understand company-specific terminology and provide consistent answers.
## MRR (Monthly Recurring Revenue)### DescriptionRefers to the recurring subscription-based revenue generated each month.Net recurring revenue reflecting new contracts, upgrades, downgrades, and churn.### CalculationMRR = Monthly subscription fee x Active subscribers### Related Metrics- ARR (Annual Recurring Revenue) = MRR x 12- Net MRR = New MRR + Expansion MRR - Contraction MRR - Churn MRR### Examples- "MRR grew 5% MoM this month"- "MRR increased ₩200M from new customer acquisition"
Keep the description to 1-2 sentences for the core, and include examples and related terms so the AI references it more accurately.
When extraction sources are configured for a category, Knowledge Graph sync auto-generates maps_to edges from the category to DB columns. This lets the agent map business terms to actual data columns.
This feature auto-converts distinct DB column values into glossary terms. Instead of manually entering hundreds or thousands of data values, build the glossary quickly by extracting directly from the DB.
Click ”+ Import from database or file” at the top-right of the glossary detail page, then click “Import from database” in the modal
2
Configure extraction
Field
Description
Database
DbSphere connection to extract values from
Table
Target table
Term column
Column to extract DISTINCT values from
Synonym column
Column to read as synonyms
Description column
Column to read as descriptions
Reference columns (LLM hint)
Additional columns to pass as LLM context
Category (required)
Category to assign to extracted terms
LLM enrichment
Toggle whether to use LLM enrichment
Category is required. It must be specified to track extraction sources and auto-generate mapping edges in the Knowledge Graph.
3
LLM enrichment settings (optional)
Toggle on LLM enrichment to auto-generate synonyms, descriptions, and examples for each value.
Option
Description
Synonyms
Enter guidance for auto-generating synonyms
Description
Enter guidance for auto-generating descriptions
Example
Enter guidance for auto-generating examples
LLM selection
Pick the LLM model to use from the dropdown
Batch size · Concurrency
Performance and parallel-processing options for efficiency
Checking each item (Synonyms / Description / Example) reveals per-item instruction fields. Provide different domain/tone/length guidance per item.
Examples:
Synonyms instruction: “Prioritize internal abbreviations and English acronyms”
Description instruction: “One sentence, understandable by non-technical readers”
Example instruction: “Format usable in finance team reports”
If empty, the system default prompt is used (same as before).
If you’ve specified synonym/description columns from the DB, LLM generation for those items is auto-disabled.
4
Run
Click “Extract” → Confirm the count of values found → Click OKExtraction runs in the background with a progress banner at the top. Identical terms already in the glossary are auto-skipped.
When you chat with an agent that has glossaries connected, the AI auto-recognizes related terms and references their definitions.
User: What's MRR?AI: MRR (Monthly Recurring Revenue) refers to monthly recurring revenue.Description:Recurring subscription-based revenue generated each month.Net recurring revenue reflecting new contracts, upgrades, downgrades, and churn.Calculation:MRR = Monthly subscription fee x Active subscribersRelated Metrics:- ARR (Annual Recurring Revenue) = MRR x 12- Net MRR Growth: Net MRR growth rate[Source: Marketing Glossary]
The AI auto-recognizes terms in the conversation without you asking explicitly.
User: What's a healthy LTV-to-CAC ratio?AI: Generally, an LTV/CAC ratio of 3:1 or higher is considered healthy.Term definitions:- CAC (Customer Acquisition Cost): Cost of acquiring a customer- LTV (Customer Lifetime Value): Lifetime value of a customer
Can the AI understand terminology without a glossary?
The AI understands common terminology, but may not accurately know company-specific terms (internal project codes, internal metric names) or recent domain-specific terms.
Use a glossary to provide accurate descriptions to the AI.
Can I connect multiple glossaries to one agent?
Yes — connect multiple glossaries to a single agent.
Connecting a marketing + finance glossary together lets the agent reference terms from both domains.
What's the difference between a Knowledge Base and a glossary?
Knowledge Base: Stores entire document content and finds related content via vector search (RAG)
Glossary: Stores only individual terms and definitions for fast reference
Use Knowledge Bases for document-based Q&A and glossaries for term reference.
How many synonyms can I register?
No limit. The more expressions you register, the better the AI recognizes terms.
I'm worried about LLM costs for DB value extraction
Turn off LLM enrichment to extract only DB values without LLM cost. If synonym/description columns are already in the DB, specifying those columns is the most economical.
What's the relationship between glossary and Knowledge Graph?
The glossary manages term definitions, while the Knowledge Graph connects terms to DB columns and documents. When extraction sources are configured on a category, the KG auto-generates term-to-column mapping edges.
