← All Workflows

🎨 Character Design Studio

An art director agent that creates consistent characters for Stable Diffusion. It remembers every character it has designed and cross-references them for visual coherence.

Nodes 7
Agents 1
Agent ID art-director@localhost
Key Feature Cross-run Memory
Remember Recall LLM ToolParser

Pipeline Flow

Config flows from Setup through the entire graph. Memory is loaded, combined with the prompt, sent to the LLM, then parsed for tool calls.

Identity
Setup
art-director@localhost
Memory
Recall
query: "character"
Prompt
SystemPrompt
art director persona
Utility
TextCombine
prompt + memories
Inference
LLM
system + user prompt
Parse
ToolParser
extract remember calls
Display
TextDisplay
final output
Agent Setup Context & Input LLM Generation Parse & Display

How It Works

📚

Memory Recall

Before generating, the agent searches its memory for previously designed characters. This ensures Elara always has red hair and arcane tattoos, no matter which session you run.

🤖

LLM + Tool Use

The LLM responds with both the visual description AND an inline [TOOL:remember] call. ToolParser executes it, saving the character to disk automatically.

🔄

Cross-Run Consistency

Queue the workflow again with a different character name. The agent loads all previous designs and contrasts them — proven in live testing with Elara + Thorne.

Node Breakdown

NodeTypeRole
SetupAgentAZAll_SetupAgent identity: art-director@localhost
RecallAgentAZAll_RecallSearch memory for "character" entries
TextCombineAgentAZAll_TextCombineMerge user prompt with recalled memories
SystemPromptAgentAZAll_SystemPromptArt director persona + tool instructions
LLMAgentAZAll_LLMGenerate character description + SD prompt
ToolParserAgentAZAll_ToolParserExecute remember tool calls, save to disk
TextDisplayAgentAZAll_TextDisplayShow clean output to user