Expand description
What an agent writes about its own run.
A run leaves three kinds of trace and they answer different questions. The RunRecord says
that it happened — agent, outcome, duration, cost — and is what a list is built from. A
HelpRequest says what got in the way. This says what the agent actually did, in its own
words, and it is the only one of the three a person reads for content rather than for status.
§Why the agent writes it rather than the Tower capturing output
A transcript is not a report. It contains everything the agent thought, including the three approaches it abandoned, and reading one to find out what happened is slower than doing the work again. The Tower could persist transcripts and separately ask for a summary, but asking an agent to state its own conclusion has a second effect worth more than the storage: an agent that must write down what it concluded is an agent that has to decide what it concluded.
§Why it is capped
Reports accumulate at one per run and are read in lists. An unbounded body is a slow page and eventually a large disk, and an agent asked for “a report” with no limit will produce its transcript. The cap is generous enough for real prose and small enough that the intent is obvious.
Structs§
- Report
- What one run concluded.
Constants§
- MAX_
ARTIFACTS - Most artefacts one report may name.
- MAX_
BODY - Longest a body may be, in characters.
- MAX_
HEADLINE - Longest a headline may be.