From 4aa41d1ddaa63ed1d92a95ae518436b50a37a95b Mon Sep 17 00:00:00 2001
From: XuhuiZhou <zhouxuhui2018@gmail.com>
Date: Fri, 2 Aug 2024 12:01:09 -0700
Subject: [PATCH] update doc

---
 docs/scenario_guideline.md | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/docs/scenario_guideline.md b/docs/scenario_guideline.md
index e1c0def..e1ae279 100644
--- a/docs/scenario_guideline.md
+++ b/docs/scenario_guideline.md
@@ -1,3 +1,7 @@
+> [!IMPORTANT]
+> Now one could add “grounding_engine_guide” key in the scenario to steer the generation of the LLM grounding engine. (e.g., “grounding_engine_guide”: “include the cleaning agent’s name (Adam Smith) in the body of the email related to cleaning services, when the AI agent uses GmailReadEmail”).
+> Now all the output of the observation is invisible to the human agents. An AI agent can use speak to report the observation back if needed.
+
 ## Scenario Guideline
 Before you start, please make sure you are in the `feature/new_scenario_across_domains` branch and you will push your changes to this branch.