#36Run totals from the traceEasyML System DesignAgentic AI
Run totals from the trace
Background
The summary footer reports three run-level totals. An event counts as an LLM call if its type is any of llm_call, plan, synthesise, or critic_verdict (all of which invoke the model); searches and attempts are counted by their own event types.
Problem statement
Implement trace_totals(events) returning (total_llm, total_searches, attempts).
Input
events— list of event dicts with an"event"key.
Output
Returns a tuple:
total_llm— events whose type is in{"llm_call", "plan", "synthesise", "critic_verdict"},total_searches— events of type"search",attempts— events of type"attempt_start".
Examples
Example 1
Input: [{"event": "plan"}, {"event": "llm_call"}, {"event": "search"},
{"event": "attempt_start"}, {"event": "critic_verdict"}]
Output: (3, 1, 1)
Explanation: LLM = plan + llm_call + critic_verdict = 3; searches = 1; attempts = 1.
Constraints
total_llmsums the four model-invoking event types.total_searchescounts"search";attemptscounts"attempt_start".
Notes
- "Total LLM calls" is usually higher than people expect — planning, synthesis, and every critique each cost a call.
Python
Loading...
▶ Run executes the 3 visible sample tests below in your browser. Submit runs the full suite — including hidden tests — on the server for an official verdict.
- •Reference example: mixed events
- •Sample: empty input yields zeros
- •Reference example: synthesise counts as an LLM call