Skip to content
LOPS Docs

Role Creator

v2.0.0New

The Role Creator generates production-ready system prompts for AI assistants. Describe the persona you want, and the operator uses a meta-prompt architecture with optional few-shot examples to craft a detailed, structured system message. The generated role can be pushed directly to an Agent LOP for immediate use.

Key Features

Section titled “Key Features”
  • Meta-prompt architecture — uses a customizable meta-prompt to instruct the LLM on how to generate high-quality system prompts
  • Few-shot examples — includes editable example personas to guide the LLM’s output style and structure
  • Quality level control — scales output from concise (2-3 paragraphs) to comprehensive (10+ paragraphs)
  • Tool-aware generation — reads tool definitions from a target Agent and incorporates tool usage guidance into the generated role
  • Direct agent integration — pushes the generated role to an Agent LOP via DAT reference or text copy

Input/Output

Section titled “Input/Output”

Inputs

Section titled “Inputs”

None. The operator is configured entirely through its parameter panel.

Outputs

Section titled “Outputs”
  • Output — the generated system prompt text
  • History — a log of all generations with timestamps, token usage, model, response time, and the full output
  • Meta Prompt — the meta-prompt template used to instruct the LLM (editable)
  • Examples — the few-shot example roles included when Use Examples is enabled (editable)
  • Info — usage instructions and tips

Use the Display menu on the Role Creator page to switch the node viewer between these outputs.

Usage Examples

Section titled “Usage Examples”

Generate a Basic Role

Section titled “Generate a Basic Role”
  1. On the Role Creator page, enter a description in the Role Description field. Be specific about capabilities, tone, and constraints — for example: “A technical support specialist for Python programming who gives patient, educational explanations to beginners.”
  2. Optionally fill in Context Info with additional details like target audience or domain.
  3. Set Quality Level to control output depth — Concise for quick roles, Comprehensive for extensive guidance with edge cases.
  4. Pulse Generate Role.
  5. The generated system prompt appears in the node viewer (set Display to Output).

Push a Role to an Agent

Section titled “Push a Role to an Agent”
  1. On the Role Creator page, set Target Agent to point to your Agent LOP.
  2. Choose a Copy Mode:
    • Reference DAT — the Agent’s system message parameter references the Role Creator’s output DAT, so any future regeneration automatically updates the Agent.
    • Copy Text — copies the generated text directly into the Agent’s existing system message DAT.
  3. Pulse Copy to Agent to push the current output, or enable Auto Copy to push automatically after every generation.

Generate Tool-Aware Roles

Section titled “Generate Tool-Aware Roles”

If your target Agent has tool LOPs connected (such as Search, Web Viewer, or custom tools):

  1. Set the Target Agent parameter to your Agent LOP.
  2. Enable Include Tool Context.
  3. Pulse Generate Role. The operator reads all tool definitions from the Agent’s tool sequence and includes them in the generation prompt, so the resulting system message contains guidance on when and how to use each tool.

Best Practices

Section titled “Best Practices”
  • Be specific in your description. “A helpful assistant” produces generic output. “A concise code reviewer for Python and TypeScript who flags security issues and suggests performance improvements” produces a much more effective role.
  • Use Context Info for audience and constraints. Separate the “what” (Role Description) from the “who and where” (Context Info) for cleaner generation.
  • Start with Balanced quality and move to Detailed or Comprehensive only when the role needs extensive edge-case handling.
  • Edit the Meta Prompt to change how roles are generated globally. Switch Display to Meta Prompt to view and edit the template directly.
  • Edit the Examples to steer output style. Switch Display to Examples to view and modify the few-shot examples. Toggle Use Examples off if you prefer zero-shot generation.
  • Use Reference DAT copy mode for iterative workflows — regenerate the role and the Agent picks up the new version automatically without manual copying.

Troubleshooting

Section titled “Troubleshooting”
  • Empty output after pulsing Generate Role — check that Role Description is not empty. The operator skips generation if the description field is blank.
  • “Could not determine API Server or Model” in the logger — verify your model configuration on the Model page. Ensure an API server is selected and a model is chosen.
  • Copy to Agent does nothing — confirm the Target Agent parameter points to a valid Agent LOP. The target must have a Systemmessagedat parameter (all Agent LOPs do).
  • Tool context section is empty — the target Agent must have tool LOPs connected and active in its tool sequence. Verify tools are wired and not disabled on the Agent.

Parameters

Section titled “Parameters”

Role Creator

Section titled “Role Creator”
Status (Status) op('role_creator').par.Status Str

Current operator status

Default:
"" (Empty String)
Active (Active) op('role_creator').par.Active Toggle

Indicates if generation is in progress

Default:
False
Generate Role (Generate) op('role_creator').par.Generate Pulse

Generate the AI role definition

Default:
False
Role Description (Roledescription) op('role_creator').par.Roledescription Str

Describe the AI role, capabilities, tone, and constraints in detail

Default:
"" (Empty String)
Context Info (Contextinfo) op('role_creator').par.Contextinfo Str

Additional context: domain expertise, use cases, target audience

Default:
"" (Empty String)
Quality Level (Qualitylevel) op('role_creator').par.Qualitylevel Menu

Controls the depth and detail of generated role

Default:
concise
Options:
concise, balanced, detailed, comprehensive
Use Examples (Useexamples) op('role_creator').par.Useexamples Toggle

Include example roles for few-shot learning

Default:
False
Target Agent (Targetagent) op('role_creator').par.Targetagent OP

Agent operator to receive the generated role

Default:
"" (Empty String)
Copy to Agent (Copytoagent) op('role_creator').par.Copytoagent Pulse

Copy generated role to target agent

Default:
False
Auto Copy (Autocopy) op('role_creator').par.Autocopy Toggle

Automatically copy to agent after generation

Default:
False
Copy Mode (Copymode) op('role_creator').par.Copymode Menu

How to transfer the role to the agent

Default:
reference
Options:
reference, text
Include Tool Context (Includetoolcontext) op('role_creator').par.Includetoolcontext Toggle

Include tool definitions from target agent to create tool-aware system prompts

Default:
False
Display (Display) op('role_creator').par.Display Menu

Select what to display in the viewer

Default:
Output
Options:
Output, History, MetaPrompt, Examples, Info
Examples DAT (Examplesdat) op('role_creator').par.Examplesdat OP

DAT containing example role definitions

Default:
"" (Empty String)
Meta Prompt DAT (Metapromptdat) op('role_creator').par.Metapromptdat OP

DAT containing the meta-prompt template

Default:
"" (Empty String)

Model

Section titled “Model”
Output Settings Header
Max Tokens (Maxtokens) op('role_creator').par.Maxtokens Int
Default:
256
Range:
0 to 1
Slider Range:
16 to 4096
Temperature (Temperature) op('role_creator').par.Temperature Float
Default:
0.0
Range:
0 to 1
Slider Range:
0 to 1
Model Selection Header
Use Model From (Modelselection) op('role_creator').par.Modelselection Menu
Default:
chattd_model
Options:
chattd_model, custom_model, controller_model
Controller [ Model ] (Modelcontroller) op('role_creator').par.Modelcontroller OP
Default:
"" (Empty String)
Select API Server (Apiserver) op('role_creator').par.Apiserver StrMenu
Default:
openrouter
Menu Options:
  • openrouter (openrouter)
  • openai (openai)
  • groq (groq)
  • ollama (ollama)
  • gemini (gemini)
  • lmstudio (lmstudio)
  • custom (custom)
  • anthropic (anthropic)
AI Model (Model) op('role_creator').par.Model StrMenu
Default:
llama-3.2-11b-vision-preview
Menu Options:
  • gpt-oss-120b (openai/gpt-oss-120b)
  • gpt-oss-20b (openai/gpt-oss-20b)
  • gpt-oss-safeguard-20b (openai/gpt-oss-safeguard-20b)
  • llama-3.1-8b-instant (llama-3.1-8b-instant)
  • llama-3.3-70b-versatile (llama-3.3-70b-versatile)
  • kimi-k2-instruct-0905 (moonshotai/kimi-k2-instruct-0905)
  • qwen3-32b (qwen/qwen3-32b)
  • allam-2-7b (allam-2-7b)
  • compound (groq/compound)
  • compound-mini (groq/compound-mini)
  • llama-4-maverick-17b-128e-instruct (meta-llama/llama-4-maverick-17b-128e-instruct)
  • llama-4-scout-17b-16e-instruct (meta-llama/llama-4-scout-17b-16e-instruct)
  • llama-guard-4-12b (meta-llama/llama-guard-4-12b)
  • llama-prompt-guard-2-22m (meta-llama/llama-prompt-guard-2-22m)
  • llama-prompt-guard-2-86m (meta-llama/llama-prompt-guard-2-86m)
Search (Search) op('role_creator').par.Search Toggle
Default:
False
Model Search (Modelsearch) op('role_creator').par.Modelsearch Str
Default:
"" (Empty String)

Provider Model Documentation

Consult the documentation for your chosen provider to find supported models, API key information, and usage limits.
View LiteLLM Supported Providers →

Changelog

Section titled “Changelog”
v2.0.02025-12-06

## Complete Rewrite

This is a complete rewrite of the role generation operator with a focus on transparency and user configurability. Previous versions are archived in custom_operators/role_creator/archive/.

Core Features

  • Meta Prompt System - Fully editable meta-prompt DAT controls all generation instructions
  • Few-Shot Examples - Editable examples DAT for customizing generation style
  • Quality Levels - Concise, Balanced, Detailed, Comprehensive output options
  • Tool Context Integration - Include agent tool definitions for tool-aware system prompts
  • Agent Integration - Direct copy to target agent with reference or text modes
  • Generation History - Full history table with complete description and output (no truncation)

Design Principles

  • Full Transparency - All generation bias is in user-editable DATs, not hidden in code
  • No Truncation - History stores complete inputs and outputs
  • Single Page UI - All parameters on one "Role Creator" page with clear sections
  • Async Generation - Non-blocking API calls via ChatTD

Parameter Layout (Role Creator page)

1. Status section - Status (readonly, expression-linked), Active

2. Actions section - Generate Role

3. Input section - Role Description, Context Info, Quality Level

4. Generation Options - Use Examples

5. Agent Integration - Target Agent, Copy to Agent, Auto Copy, Copy Mode, Include Tool Context

6. Display section - Output/History/MetaPrompt/Examples/Info viewer selector

7. Advanced section - Examples DAT, Meta Prompt DAT references

Editable DATs

  • meta_prompt_dat - Complete control over generation instructions
  • examples_dat - Few-shot examples for style guidance
  • info_dat - Help text
  • output_dat - Generated role output
Section titled “Related Operators”