How Context Connections Make Your Agent Smarter

Context Connections are links between related pieces of your knowledge base. They help your agent retrieve relevant information together, instead of finding facts in isolation.

The Problem Connections Solve

Without connections, your agent searches the entire knowledge base for every question. A caller asking about scheduling might get a mix of scheduling facts, billing FAQs, and general information — some relevant, some not.

With connections, related knowledge clusters together. Your scheduling-related facts, FAQs, policies, and intents form a group. When the conversation is about scheduling, the agent draws from that cluster first.

How Connections Work

A connection links a source item to a target item with a relationship type and strength:

• Source: "schedule_appointment" intent
• Target: "Cancellation Policy" policy
• Connection type: related_knowledge
• Strength: 0.8 (strong)

This tells the system: when the conversation involves appointment scheduling, the cancellation policy is strongly relevant. The agent is more likely to surface it proactively or find it quickly when the caller asks.

Types of Connections

• Intent → Knowledge — link intents to the facts, FAQs, and policies the agent needs for that call type. The "insurance_question" intent connects to your insurance FAQ, accepted providers fact, and billing policy.
• Knowledge → Knowledge — link related knowledge items. Your parking fact connects to your address fact and your "what to bring to your first visit" FAQ.
• Keyword clusters — keywords extracted from your knowledge base create implicit connections. Items sharing keywords like "appointment", "scheduling", "booking" naturally cluster together.

Why This Matters

For small knowledge bases (under a few hundred items), connections add helpful structure but aren't critical — the agent can search everything quickly.

For larger knowledge bases, connections become important for precision. Without them, a search for "cancellation" might return your cancellation policy, a FAQ about event cancellations, and a fact about cancelled insurance — all valid matches but only one is relevant to what the caller is asking about. Connections help the agent prioritize the right context.

Automatic vs Manual Connections

When your agent is first configured, connections are generated automatically based on content similarity and keyword overlap. As you add knowledge, new connections are suggested. You can also create manual connections from the dashboard to explicitly link items you know belong together.