D&D Vault

Entity-Update-Workflow

7 min read

Entity Update Workflow

Purpose

This document outlines the process for updating Entity pages (NPCs, PCs, Items, Locations) using the Automated Architect Pipeline. It relies on the generate_pages.py script to perform non-destructive updates to your vault.

1. Prerequisites

Ensure you have these three components ready:

  1. The Script: generate_pages.py (Version 3) in your root folder.
  2. The Prompt: System/Prompts/Architect-Prompt.md (The Generic Version).
  3. The Templates: NPC.md, Item.md, Ongoing Thread.md (Standardized with {{keys}}).

2. The Workflows (Automated)

Workflow A: The “Sniper” Update (Backfilling History)

Best for: Updating an existing NPC/Location with missed events from past sessions. Concept: Use Obsidian’s “Linked Mentions” to find relevant entries, then process them in a batch.

  1. Identify Sources: Open the target note (e.g., Corbin.md) and check Backlinks. Note which Entries mention him (e.g., Entry 37, Entry 38).
  2. Gather Context: Open a chat and upload:
    • The Target File (Corbin.md).
    • The Source Chronicles (Entry 37.md, Entry 38.md).
  3. Prompt: Paste the Architect Prompt.
    • No extra typing needed. The prompt automatically detects the target and sources.
  4. Execute:
    • Copy the JSON output to update.json.
    • Run: python generate_pages.py.
    • Verify: Check git diff to ensure history was appended correctly.

Workflow B: The “New Session” Routine (Day-to-Day)

Best for: Processing the latest session (Entry 39) immediately after play.

  1. Preparation:
    • Run contextbomb in your terminal to refresh ~/Vault_Context.txt.
  2. Scout:
    • Open a new chat.
    • Upload Entry XX.md AND Vault_Context.txt.
    • Paste the Scout Prompt.
    • Result: A checklist of “Existing Files” and “New Files”.
  3. Architect:
    • Open a new chat (to clear the heavy context).
    • Upload the Target Files identified by the Scout (e.g., Corbin.md, Onasea.md) plus Entry XX.md.
    • Paste the Architect Prompt.
  4. Execute: Paste JSON to update.json, run script.

Workflow C: The PC “Highlight Reel” (Main Characters)

Best for: Gage, Lavender, Theren. Concept: We do not log every entry for PCs. We only log major character beats.

  1. Target: Update Gage/Gage.md (The Bio Module), not the Dashboard.
  2. Prompt: Use the Architect Prompt, but upload only the specific “Big Moment” entry.
  3. Outcome: The script appends a bullet to ## Biography or ## Notable Exploits.
    • Example: “Entry 38: Led the underwater escape.”

3. Maintenance & Cleanliness

The “Header Rule” (Critical)

The Python script looks for specific headers to append text. To ensure automation works, your existing files must use these standard headers:

File TypeStandard HeaderScript Behavior
NPC## HistoryAppends narrative beats here.
Location## HistoryAppends events that happened here.
Item## LoreAppends new owners or discoveries.
PC (Bio)## BiographyAppends major life events.

Fix: If you see ## Campaign Notes, rename it to ## History manually before running the script.

The “Git Safety Net”

Always commit before running the script.

git add .
git commit -m "Pre-update state"
python generate_pages.py
# Check diffs. If good:
git commit -am "Automated update from Entry 38"

4. Manual & Research Methods

Method D: The “Grep Dump” (Deep Research)

Best for: Fact-checking a minor detail across 50 files without updating them.

  1. Create Context File: Use grep with context (-C) to capture the story around a keyword.

    grep -C 20 "Ring of Orris" content/Campaigns/Campaign\ 1/Chronicles/*.md > Context.txt

  2. Action: Upload Context.txt to the LLM to answer specific questions (“When did we last use the Ring?”).

Method E: The “Context Bomb” (Full Arc Backfill)

Best for: Backfilling a main character with 20+ sessions of history at once.

  1. Preparation:
    • Run contextbomb in your terminal to generate ~/Vault_Context.txt.
  2. The Prompt:
    • Open a new chat.
    • Upload ~/Vault_Context.txt.
    • Copy/Paste the content of Context-Bomb-Prompt.md.
    • (Optional: Edit the “Target Character” line if backfilling someone else).
  3. Execution:
    • Copy the markdown block from the LLM response.
    • Paste it manually into the character’s Biography.md or index.md.

Best for: Ensuring your manually formatted Chronicle Entry has the correct links before you finalize it.

  1. Preparation: Run contextbomb (alias for ~/scripts/generate_context.sh) to refresh the vault index.
  2. Prompt:
    • Open a new chat. Upload Vault_Context.txt + Your formatted Entry XX.md.
    • Paste the Wikilink Spotter Prompt (System/Prompts/Wikilink-Spotter.md).
  3. Action:
    • The LLM provides a checklist of proper nouns found in your vault (e.g., “Link ‘Corbin’ in Section 2”).
    • You manually bracket [[Corbin]] in Obsidian to preserve your specific formatting.

Workflow G: The “Profile Builder” (NPC Overhaul)

Best for: Generating a full wiki page for a specific NPC (e.g., Anaphel) by analysing their entire history to determine if they are a “Major” or “Minor” character.

  1. Preparation:
    • Run contextbomb (alias for ~/scripts/generate_context.sh) to ensure the Archives are up to date.
  2. The Chat Setup:
    • Open a new chat.
    • Upload: Vault_Context.txt (The History).
    • Upload: Templates/NPC.md (The Structure).
    • Upload (Optional): The existing file (e.g., Anaphel.md) if you want to preserve current metadata.
  3. The Prompt:
    • Paste the Grand Biographer Prompt (System/Prompts/Grand-Biographer.md).
    • Edit the Target: Change [[TARGET_NAME]] to the specific NPC (e.g., [[Anaphel]]).
  4. The Result:
    • The LLM will assess the NPC’s importance based on their history.
    • It produces a single Markdown code block (either the full “Gage Standard” for major NPCs or a summarized version for minor ones).
    • Copy/Paste this block into the NPC’s markdown file.

Workflow H: The “Chronicle Transformation” (Post-Session)

Best for: Converting raw narrative logs from Discord into vault-ready entries for Quartz. Concept: Structural cleanup and wikilink integration for consistency.

  1. Identify Source: Download the raw session narrative (e.g., DnD 39.docx).
  2. Chat Setup:
    • Open a new chat.
    • Upload: The raw narrative file.
  3. Prompt: Paste the Chronicle Architect Prompt (System/Prompts/Chronicle-Architect.md).
  4. Execute:
    • Copy the resulting Markdown into a new file: Campaigns/Campaign 1/Chronicles/Entry XX.md.
    • Snapshot Update: Copy the bullet points from Phase 4 of the LLM response and paste them into the “SECTION 0” block of your generate_context.sh script to keep your “Now” data current.

Workflow I: The “Lore Forge” (New Entity Creation)

Best for: Creating dedicated wiki pages for new NPCs, Items, or Locations discovered during a session (e.g., Captain Voss, The Tidecaller). Concept: Maintaining standard schema across all lore pages using your templates.

  1. Identify Source: Use a finalized Chronicle Entry (e.g., Entry 39.md).
  2. Preparation: Run contextbomb to ensure the latest templates from Section 1.5 are in your export file.
  3. Chat Setup:
    • Open a new chat.
    • Upload: Vault_Context.txt (This provides templates and the manifest).
    • Upload: The specific source Chronicle Entry where the entity appears.
  4. Prompt: Paste the Entity Extractor Prompt (System/Prompts/Entity-Extractor.md).
  5. Execute:
    • Create the new file (e.g., NPCs/Captain Voss.md).
    • Verify: Ensure the LLM matched the YAML and heading structure from your System/Templates.

Workflow J: The “Vault Audit” (Lore Cleanup)

Best for: Ensuring no NPCs or Items from the Chronicles have been “lost” without dedicated files. Concept: Cross-referencing the Narrative Source against the File Manifest.

  1. Preparation: Run contextbomb to generate the latest index of all existing files.
  2. Chat Setup:
    • Open a new chat.
    • Upload: Vault_Context.txt.
  3. Prompt: Paste the Vault Auditor Prompt (System/Prompts/Vault-Auditor.md).
  4. The Result:
    • A list of “Untracked Entities” (People/Items mentioned in Chronicles but missing from the Manifest).
    • A list of “Dead Links” (Proper nouns that should be wikilinked but aren’t).
  5. Action: Use Workflow I to “Forge” the missing pages identified by the Auditor.

Grimmora Prompt Index

Prompt NameFile PathPrimary TriggerBest Workflow
Chronicle ArchitectSystem/Prompts/Chronicle-Architect.mdNew narrative log from Discord.Workflow H: Raw text to Obsidian Entry.
Entity ExtractorSystem/Prompts/Entity-Extractor.mdA new NPC, Item, or Location is mentioned.Workflow I: Entry to dedicated Lore Page.
Architect PromptSystem/Prompts/Architect-Prompt.mdNeed to append new events to existing files.Workflows A, B, C: Automated updates.
Scout PromptSystem/Prompts/Scout-Prompt.mdPreparing to process a session entry.Workflow B: Identifies what to update.
Vault AuditorSystem/Prompts/Vault-Auditor.mdVault feels messy or links are missing.Workflow J: Narrative vs. Manifest audit.
Grand BiographerSystem/Prompts/Grand-Biographer.mdMajor character needs a full history overhaul.Workflow G: Comprehensive profile rebuild.
Context BombSystem/Prompts/Context-Bomb-Prompt.mdNeed to catch an LLM up on a character’s arc.Method E: Full history backfill.
Wikilink SpotterSystem/Prompts/Wikilink-Spotter.mdFinalizing a manually written Entry.Method F: Proper noun highlighting.

Graph View