An assistant must look up live order status during a conversation. What do you add?
Why A is the answer
A tool the agent can call, backed by the order system
Live data needs a tool call. A vector index holds a snapshot of documents.
In detail
- Order status changes all the time.
- A vector index holds a snapshot of documents, while a tool reads the live system.
- Validate what the tool receives, and limit it to the current customer.
Use case
An illustrative scenario, not a real incident.
A customer asks where their order is. No document holds that, because it changes by the hour. The agent calls a tool backed by the order system and gets the current status.
See it in different ways
flowchart TD
A{"What kind of knowledge?"} -->|"Policies and manuals"| B["Vector Search"]
A -->|"Facts for one order or customer"| C["Lookup by key in a table"]
A -->|"Live values"| D["A tool call"]sequenceDiagram
participant U as User
participant A as Agent
participant I as Vector index
participant T as Tool
U->>A: Question
alt General policy
A->>I: Search documents
I-->>A: Passages
else Live or per-record fact
A->>T: Call with a key
T-->>A: Current value
end
A-->>U: AnswersequenceDiagram participant P0 as Customer participant P1 as Agent participant P2 as Order tool participant P3 as Order system P0->>P1: Where is my order? P1->>P2: Look up this order P2->>P3: Current status P3-->>P1: Shipped today P1-->>P0: Answer
Key point: Live data needs a tool, not an index.
Why the other options are wrong
- B. More documents added to the vector index
- More documents give a larger snapshot, and it is still not live.
- C. A larger embedding model for the order data
- A better embedding does not make data current.
- D. A longer chunk size on the order history
- Chunk size does not change how old the data is.