Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/yocxy2/2a/llms.txt

Use this file to discover all available pages before exploring further.

The Knowledge Base Manager is the administrative interface for tuning how the hybrid RAG pipeline retrieves and ranks support articles. Rather than treating all knowledge base articles equally, the system assigns each article an importance weight — a numeric multiplier that influences how prominently an article surfaces when the AI is searching for context to answer a support ticket. Agents can view all articles, inspect their current importance values, and update them directly in the UI. The page is built in src/pages/KnowledgeBaseManager.tsx and uses the ticketService from src/services/api.ts for data operations.

​
Accessing it

1

Open the application

Navigate to http://localhost in your browser (or http://localhost:3000 in development mode).
2

Click Knowledge Base

In the blue top navigation bar, click Knowledge Base. This routes to /knowledge-base.

​
Importance weight system

Each knowledge base article carries an importance field — a float between 0.0 and 2.0. This value acts as a multiplier within the hybrid RAG scoring formula that determines which articles are retrieved and passed as context to GPT-4o-mini. The hybrid scoring formula is: 
finalScore = (vectorSimilarity × 0.8) + (recencyScore × 0.2) × importance
The importance value scales the recency component of the score. An article with importance: 2.0 will have its recency contribution doubled, making it much more likely to appear in the top retrieved results even when it is slightly older or the vector similarity is not the highest. Key thresholds:
  • Articles with importance ≥ 1.5 are treated as critical and are consistently prioritized in retrieval.
  • The default importance for new articles is 1.0.
  • Setting importance to 0.0 effectively removes the recency bonus entirely — the article can still be retrieved, but only on strong vector similarity.

​
Color coding

Each article’s importance value is rendered as a color-coded badge so agents can quickly scan the list for articles that need attention:

Critical

Purple — Importance ≥ 1.5. Highest priority articles. These surface in nearly every relevant query. Use for foundational policies, high-frequency issues, or legally important information.

Standard

Green — Importance ≥ 1.0. Normal priority. The default level for most articles. Retrieved regularly when the vector similarity is a good match.

Low

Yellow — Importance ≥ 0.5. Deprioritized. These articles are surfaced less frequently. Useful for niche topics that should not compete with core articles.

Minimal

Gray — Importance < 0.5. Rarely surfaced. The recency bonus is nearly eliminated. Reserve this for outdated articles that are being kept for archival purposes.
The color logic is implemented in KnowledgeBaseManager.tsx: 
const getImportanceColor = (importance: number): string => {
  if (importance >= 1.5) return 'text-purple-600 bg-purple-100';
  if (importance >= 1.0) return 'text-green-600 bg-green-100';
  if (importance >= 0.5) return 'text-yellow-600 bg-yellow-100';
  return 'text-gray-600 bg-gray-100';
};

​
Editing importance

Adjusting an article’s importance weight is done inline on the Knowledge Base Manager page:
1

Find the article

Scroll through the article list to find the entry you want to update. Each card shows the article title, category tag, content preview, and its current importance badge.
2

Click Edit

Click the blue Edit button on the right side of the article card. The importance badge is replaced with a numeric input field pre-filled with the article’s current value.
3

Adjust the value

Use the numeric input to set the new importance. The input accepts values from 0 to 2 in increments of 0.1.
<input
  type="number"
  step="0.1"
  min="0"
  max="2"
  value={editImportance}
  onChange={(e) => setEditImportance(parseFloat(e.target.value))}
/>
4

Save or cancel

Click Save to apply the change, or Cancel to discard it and return to the badge view. After saving, the importance badge immediately reflects the new value with the appropriate color.

​
Default articles

The system is preloaded with 3 knowledge base articles that cover the most common support categories. These are loaded into local state when the Knowledge Base Manager page mounts:
#TitleCategoryDefault Importance
1How to request a refundReturns1.0
2Payment methods acceptedBilling1.0
3Track your orderShipping0.8
These articles form the base retrieval set. For new deployments, start by reviewing whether any of these should be elevated to importance: 1.5 based on your support ticket volume by category.
Changes made in the Knowledge Base Manager UI are client-side only in the current implementation. The updated importance values are written to local React state (useState) and are not persisted to the database. If you restart the page, all values reset to their defaults.To persist importance changes, either call the addKnowledgeArticle function directly in the RAG service backend, or extend the API with a PATCH /api/v1/knowledge/:id endpoint and wire it to the handleUpdateImportance function in KnowledgeBaseManager.tsx.

Build docs developers (and LLMs) love

Get started for freeTalk to us