Agents And Tool Use easy
You are building a support agent with a 'lookup_order' tool. A teammate asks where the database query actually runs when the model 'calls' the tool. Which description is correct?
- A The model connects to the database directly using credentials embedded in the system prompt.
- B The model provider's servers run the query and stream the rows back into the response.
- C The model emits a structured request; your host code runs the query and returns the result.
- D The tool schema compiles into a stored procedure that the model triggers at inference time.
Answer: C. The model emits a structured request; your host code runs the query and returns the result.
A **tool call** is only structured output: the model proposes a function name and arguments, and the **host** decides whether to execute it and returns the result. The tradeoff is that the host must validate, authorize, and run every call, but this is what keeps credentials and side effects under your control. Putting credentials in the prompt (A) assumes the model executes code; it never does.
One system design question a day, with spaced repetition so the ones you miss come back.
Get the appRelated questions
- You are defining a 'schedule_meeting' tool for an agent. Which schema decision most reduces wrong or malformed calls?
- A model emits three parallel tool calls: `get_inventory`, `get_price`, and `get_shipping`. To cut latency, the host appends each result to the transcript as it arrives and immediately re-invokes the model. Runs start failing with malformed-request errors, and some produce hallucinated shipping costs. What should the host do?
- A team splits a coding agent into planner, coder, tester, and reviewer agents, and each receives the full transcript of the previous agent. The expected quality gains did not appear. Cost tripled, and the reviewer often fixates on approaches that earlier agents abandoned. When is a multi-agent split actually justified?
- A support agent should remember each customer's preferences across months of sessions. The team appends every past session summary to the system prompt. After six months, prompts exceed 80,000 tokens, latency has doubled, and the agent sometimes ships to an address the customer replaced weeks ago. What redesign fixes both problems?
- A teammate proposes splitting a single coding agent into five agents, arguing that more agents means smarter results. When is a multi-agent split actually justified?
- A data-analysis agent runs model-generated Python inside the same container as your API service. That container has database admin credentials in environment variables. A security review flags this as critical. What is the best mitigation?