Refactor the dnd scene_map prompt to use the common cached system message and transcript

prompts/pipeline-dnd/analyze/scene_map/scene_map.instructions.md

@@ -1,22 +1,4 @@
-Generate a structured scene map for the provided D&D session transcript.
-
-This is an intermediate artifact for an automated analysis pipeline. Your task is only to divide the transcript into coherent scenes. Do not produce a session recap, character analysis, tactical analysis, continuity analysis, or recommendations.
-
-Use the transcript as the source of truth. Optional metadata, character roster, and glossary are provided only to help identify speakers, characters, places, factions, spells, items, and proper nouns.
-
-Return only valid JSON matching the provided JSON Schema.
-
-Use the following party roster to resolve speakers and character references:
-
-- Mike Brown: DM
-- Eric Rakestraw: Player (Hrank Greenfield, Nature Cleric of Chauntea)
-- Adam Rakestraw: Player (Alan Hohey, Druid)
-- Zach Anthony: Player (Svend Balderk, Barbarian)
-- Clint Twellman: Player (Popov, Paladin)
-
-Task:
-
-Divide the transcript into sequential, contiguous, non-overlapping scenes.
+INSTRUCTIONS: Your task is to divide the provided transcript into sequential, contiguous, non-overlapping scenes.
 
 A scene is a coherent unit of play. A new scene should usually begin when there is a meaningful change in location, objective, threat, activity, encounter, or mode of play.
 
@@ -61,44 +43,16 @@ For each scene:
 
 Primary mode guidance:
 
-Use Recap for opening recap, initiative setup, session framing, or immediate continuation from prior events.
-
-Use Combat when active combat or combat-resolution mechanics dominate.
-
-Use Social when conversation, negotiation, persuasion, interrogation, or NPC interaction dominates.
-
-Use Exploration when the party is investigating, scouting, navigating, searching, or probing the environment.
-
-Use Travel when movement between locations is the main activity.
-
-Use Planning when the party is primarily discussing options or choosing a course of action.
-
-Use Preparation when the party is actively preparing for an upcoming combat, journey, encounter, or other event.
-
-Use Downtime when the party is engaged in downtime activities.
-
-Use Rest when resting, healing, resource recovery, or post-encounter recovery dominates.
-
-Use Rules only when table-level rules, mechanics, or adjudication dominate enough to form their own scene.
-
-Use Mixed only when no single mode clearly dominates.
+- Use Recap for opening recap, initiative setup, session framing, or immediate continuation from prior events.
+- Use Discussion when the party is primarily discussing options or choosing a course of action.
+- Use Combat when active combat or combat-resolution mechanics dominate.
+- Use Narrative for all other types of non-combat gameplay, including exploration, social interactions, shopping, preparing for battle, downtime, and so on.
 
 Session scope requirements:
 
-In session_scope.source_description, briefly describe the transcript reviewed.
-
-In session_scope.first_segment_id, provide the first segment ID covered by the scene map.
-
-In session_scope.last_segment_id, provide the last segment ID covered by the scene map.
-
-In session_scope.limitations, list any limitations that affect scene segmentation, such as:
-- missing prior-session context
-- unclear speaker attribution
-- overlapping speech
-- map-dependent positioning
-- visual tabletop references not visible in the transcript
-- ambiguous transitions
-- uncertain proper nouns
+- In session_scope.source_description, briefly describe the transcript reviewed.
+- In session_scope.first_segment_id, provide the first segment ID covered by the scene map.
+- In session_scope.last_segment_id, provide the last segment ID covered by the scene map.
 
 Boundary caveats:
 

prompts/pipeline-dnd/analyze/scene_map/scene_map.system.md

@@ -1,7 +0,0 @@
-You are a careful transcript segmentation assistant.
-
-Your job is to divide a D&D session transcript into coherent scenes for downstream analysis.
-
-Follow the user’s instructions exactly. Use only the provided transcript and metadata. Do not invent facts, infer campaign history, analyze character performance, or produce a narrative recap.
-
-Return only valid JSON that conforms to the requested schema. Do not include Markdown, comments, explanations, or text outside the JSON object.

prompts/pipeline-dnd/analyze/scene_map/scene_map.task.md

@@ -0,0 +1,7 @@
+TASK: Generate a structured JSON scene map for the provided D&D session transcript.
+
+This is an intermediate artifact for an automated analysis pipeline. Your task is to divide the transcript into sequential, contiguous, non-overlapping scenes.
+
+Return only valid JSON matching the provided JSON Schema.
+
+Next, you will be provided with detailed instructions with respect to preparation of the strucured JSON scene map.

prompts/pipeline-dnd/analyze/scene_map/scene_map.yml

@@ -1,6 +1,6 @@
 id: dnd.scene_map
 version: "1.0.0"
-default_profile: gemma-4-31b
+default_profile: nemotron-3-ultra
 description: Generates a D&D session scene map, using a transcript as input.
 inputs:
   - name: transcript
@@ -9,18 +9,18 @@ inputs:
     description: Structured recap source transcript
 messages:
   - role: system
-    content_file: ./scene_map.system.md
-  - role: user
-    content_file: ./scene_map.user.md
+    content_file: ../common/transcript.system.md
   - role: user
     content: |
       <<<CURRENT_SESSION_TRANSCRIPT
-
       {{input "transcript"}}
-
       CURRENT_SESSION_TRANSCRIPT>>>
+  - role: user
+    content_file: ./scene_map.task.md
+  - role: user
+    content_file: ./scene_map.instructions.md
 output:
   format: json
   validation_mode: json_schema
   schema_path: automation-pipeline-dnd/analyze/scene_map.schema.json
-  repair_attempts: 0
+  repair_attempts: 2