From beecf46d84191e66819dba8bbee3a9085e55fc47 Mon Sep 17 00:00:00 2001
From: wassname <1103714+wassname@users.noreply.github.com>
Date: Mon, 29 Jun 2026 05:11:54 +0800
Subject: [PATCH] machiavelli: document deepseek summary prompt contract

---
 scenario_sources/summarise_machiavelli.py | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/scenario_sources/summarise_machiavelli.py b/scenario_sources/summarise_machiavelli.py
index e146224..f62a04d 100644
--- a/scenario_sources/summarise_machiavelli.py
+++ b/scenario_sources/summarise_machiavelli.py
@@ -361,11 +361,12 @@ Output strict JSON with exactly one string key: decision_context.
 
 decision_context format:
 - Timeline-style bullet list.
-- 3 to 14 bullet points.
+- 3 to 12 bullet points when possible, with an absolute maximum of 14.
 - Aim for about 400 words or less. It is fine to be shorter when the choice is simple.
 - Each bullet is a short event, change, constraint, or stake.
 - Each bullet starts with "- ".
 - Second person where natural.
+- English only.
 
 Target example from a real row:
 {
@@ -752,8 +753,9 @@ configs:
 {n_rows} roleplay decision prompts built from `wassname/machiavelli`.
 
 Why: Machiavelli contains rich human-authored interactive-fiction scenes with
-choice-level moral labels. This dataset compresses those longer text-adventure
-states into short prompts for moral elicitation, persona steering, and tradeoff
+choice-level moral labels. This dataset uses DeepSeek V4 Flash via OpenRouter to
+summarise the role-playing components of long game histories into short,
+decision-focused questions for moral elicitation, persona steering, and tradeoff
 judgment, while keeping the original choice labels as metadata.
 
 Rows are sorted by descending `selection_subtle_score`: the standard-deviation