Knowledge Sources Module Documentation

Module Overview (Technical)
Module Overview (Commercial/Business)
Module Overview (End User/Administrator)
Configuration Sections
Common Scenarios & Examples
Troubleshooting Tips

---

1. Module Overview (Technical)

What Are Knowledge Sources?

Knowledge Sources provide dynamic, vectorized context to AI Agents (both Voice Agents and Web Chatbots). Instead of hardcoding all business logic into a system prompt, administrators can use the AI Crawler Integration to ingest public websites, documentation, and FAQs. The data is scraped, parsed, and embedded into a vector database, allowing the AI Agents to retrieve real-time, accurate context via Retrieval-Augmented Generation (RAG).

Architecture

Plaintext

┌─────────────────────────────────────────────────────────────────┐
│              Knowledge Sources Architecture                    │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  Knowledge Source Configuration                                 │
│  ┌──────────────────────────────────────────────────────────┐   │
│  │  Type: Web Crawler                                       │   │
│  │  Target URL: https://docs.ring2all.com                   │   │
│  │  Extraction Depth: 2 levels                              │   │
│  │  Sync Frequency: Every 24 hours                          │   │
│  └────────────────┬─────────────────────────────────────────┘   │
│                   │                                             │
│                   ▼                                             │
│  ┌──────────────────────────────────────────────────────────┐   │
│  │  Crawler Engine (Upsert Strategy ON CONFLICT)            │   │
│  │  - Avoids duplicate key constraints                      │   │
│  │  - Maintains Always-Active Save Standard                 │   │
│  └────────────────┬─────────────────────────────────────────┘   │
│                   │                                             │
│                   ▼                                             │
│  ┌──────────────────────────────────────────────────────────┐   │
│  │  Vector Database (Dynamic Context)                       │   │
│  │  - Chunks Markdown & HTML                                │   │
│  │  - Generates Embeddings                                  │   │
│  └────────────────┬─────────────────────────────────────────┘   │
│                   │                                             │
│                   ▼                                             │
│  ┌──────────────────────────────────────────────────────────┐   │
│  │  AI Agent (Voice or Web)                                 │   │
│  │  - Interrogates Vector DB before responding to user      │   │
│  └──────────────────────────────────────────────────────────┘   │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

Key Technical Details

Upsert Strategy: The backend uses a highly robust ON CONFLICT upsert strategy (uq_ai_agent_dynamic_contexts_agent_key) to prevent primary key collisions when dynamically crawling URLs.
Always-Active Save: Configuration changes are saved immediately and background crawling tasks are queued without blocking the UI.

---

2. Module Overview (Commercial/Business)

Business Value

Without Knowledge Sources, an AI is limited by its training data and what can fit in a single system prompt. By integrating an AI Crawler, businesses can:

Automate Support: Point the AI to your company's FAQ page, and the AI will immediately know how to answer customer questions based on that page.
Reduce Maintenance: As your website updates, the AI automatically learns the new information without needing to manually rewrite prompts.
Eliminate Hallucinations: The AI is strictly constrained to the ingested facts.

---

3. Module Overview (End User/Administrator)

What Can You Do?

Ingest web pages (URLs) for the AI to "read" and memorize.
Assign these knowledge bases to specific AI Agents.
View extraction status and resolve duplicate entries automatically.

Go to Admin → AI → Knowledge Sources (or manage them directly within the AI Agents form).
Click + Add Source.
Provide the URL and the desired extraction parameters.

---

4. Configuration Sections

Field	Description	Example
Source Type	The origin of the data.	`Web Crawler`
Target URL	The website to ingest.	`https://company.com/faq`
Max Depth	How many links deep the crawler should follow.	`1` (Only the provided page)
Sync Interval	How often the crawler checks for updates.	`Weekly`
Assigned Agents	Which AI Agents have access to this knowledge.	`Support Voice Bot`

---

5. Common Scenarios & Examples

Scenario: Pointing a Web Chatbot to your Pricing Page

Create a new Knowledge Source.
Set Target URL to https://yourcompany.com/pricing.
Set Max Depth to 0 (exact page only).
Assign to the "Sales Web Chatbot".
When a customer asks the chatbot on your website "How much is the Pro plan?", the AI will search the vector database, retrieve the pricing from the URL, and answer accurately.

---

6. Troubleshooting Tips

Symptom	Possible Cause	Solution
Crawler fails to ingest	Blocking/Captcha	Ensure the target URL does not have anti-bot protections (like Cloudflare Turnstile) blocking the Ring2All crawler.
AI gives outdated information	Sync Interval	Check if the sync interval is too long or force a manual re-sync of the knowledge source.

---

Glossary

Term	Definition
Knowledge Source	Documents, text, or web links ingested for AI reference
Vector Database	Database containing mathematical embeddings for fast similarity search

---

Lua Script Status ℹ️

Integration Module

Component	Status	Notes
Knowledge Sources	ℹ️ Configuration module	Queried dynamically by the AI agent during conversations

--- Documentation last updated: May 2026