Knowledge Base

Give your AI agents deep domain knowledge so they can answer product-specific questions, follow company policies, and provide accurate information during calls.

How It Works

Levrage.AI agents use a knowledge base to ground their responses in facts you provide. Instead of relying solely on their training data, agents reference your knowledge base to answer questions accurately and consistently.

There are two ways to provide knowledge:

Typed Knowledge

Pass knowledge directly when creating or updating an agent:

1
import requests
2
3
response = requests.post(
4
    "https://api.levrage.ai/v1/agents",
5
    headers={"Authorization": f"Bearer {API_KEY}"},
6
    json={
7
        "name": "Product Support Agent",
8
        "agent_type": "support",
9
        "language": "English",
10
        "voice": "female",
11
        "start_flow": "Hello! How can I help you today?",
12
        "end_flow": "Thanks for calling! Have a great day.",
13
        "typed_knowledge": """
14
COMPANY: Acme Electronics
15
RETURN POLICY: 30-day returns on all products. Items must be unused and in original packaging.
16
WARRANTY: 1-year manufacturer warranty on all electronics. Extended warranty available for $49/year.
17
18
PRODUCTS:
19
- SmartWatch Pro: ₹12,999 — GPS, heart rate, 7-day battery
20
- AirPods Ultra: ₹8,499 — Active noise cancellation, 24hr battery
21
- PowerBank 20K: ₹2,499 — 20,000mAh, fast charging, USB-C
22
23
STORE HOURS: Mon-Sat 10AM-8PM, Sun 11AM-6PM
24
SUPPORT EMAIL: support@acme-electronics.com
25
SUPPORT PHONE: +91 1800-123-4567 (toll-free)
26
        """
27
    }
28
)

Updating Knowledge

Update an agent’s knowledge base at any time:

$
curl -X PATCH https://api.levrage.ai/v1/agents/{agent_id} \
>
  -H "Authorization: Bearer $API_KEY" \
>
  -H "Content-Type: application/json" \
>
  -d '{
>
    "typed_knowledge": "Updated product catalog and policies..."
>
  }'

Writing Effective Knowledge

Do

• Structure with clear headings — Use labels like POLICY:, PRODUCT:, FAQ:
• Include specific numbers — Prices, dates, limits, measurements
• Cover common questions — Think about what customers will ask
• Keep it factual — Avoid opinions or marketing language
• Update regularly — Stale knowledge leads to wrong answers

Don’t

• Don’t paste entire websites — Curate the most relevant information
• Don’t include internal jargon — Agents will repeat what you give them
• Don’t add contradictory info — The agent may get confused
• Don’t write instructions here — Use the prompt field for behavior instructions

Example: Real Estate Agent Knowledge

COMPANY: Dream Homes Real Estate
LOCATION: Mumbai, Maharashtra
AVAILABLE PROPERTIES:
1. Sea View Apartment, Worli
   - 3 BHK, 1,450 sq ft
   - Price: ₹3.2 Cr
   - Amenities: Pool, gym, parking, 24/7 security
   - Ready to move
2. Garden Villa, Powai
   - 4 BHK, 2,800 sq ft
   - Price: ₹5.5 Cr
   - Amenities: Private garden, clubhouse, EV charging
   - Possession: June 2026
PAYMENT: EMI options available via HDFC, ICICI, SBI
SITE VISITS: Available Mon-Sun, 10AM-6PM
CONTACT: Rahul Mehta (Senior Sales Manager) — +91 98765 43210

Knowledge Base via Dashboard

For large documents (PDFs, manuals, catalogs), upload them through the Dashboard:

1. Go to Levrage.AI Studio
2. Select your agent → Knowledge Base tab
3. Click Upload Document
4. Supported formats: PDF, TXT, DOCX (max 10MB)

The uploaded documents are automatically processed and made available to your agent during calls.

How Agents Use Knowledge

During a call, when the user asks a question, the agent:

1. Searches the knowledge base for relevant information
2. Extracts the most relevant facts
3. Formulates a natural response using those facts
4. Cites specific details (prices, names, policies) accurately