What is the difference between the “role” and the “reminder” text fields ina pickaxe settings? Isn’t it just system prompt to inject automatically into the every message?
Hi! Great question.
The Role Prompt is the main instruction that powers your AI app. It sets the overall behavior and personality for your chatbot or tool.
The Model Reminder is a separate field where you can add an extra instruction that gets prepended to every user message. This is useful for special rules or style guidelines you want the bot to remember throughout the conversation. While both are types of system prompts, the reminder specifically acts as a consistent nudge for the AI with each interaction.
The reminder can be used to reinforce key elements such as tone, formatting, or specific instructions that the bot should consistently keep in mind.
So if I got it right, the role promt injected only once at very beginning of the communication and staying there for the whole chat. Meaning it will start degrading when the context will become longer?
Hey @conrad the Role section is intended for the system prompt or Core prompt of your Pickaxe. It’s like an SOP for how the Pickaxe should run.
The ‘Model reminder’ section is intended to over-ride the underlying LLMs safety settings (to an extent).
For example, I was building a Pickaxe for a client who is a certified electrician in his country. He hired me to build a Pickaxe that is intended for field use by other certified electricians. During beta testing, we discovered that the LLM kept on saying to his users “…always consider hiring a professional electrician to avoid electrical hazards…”
So I added an over-ride prompt informing the LLM of the user’ profile and instructed it to refrain from telling professional users to hire other professionals and that solved the problem.
Thank you for the detailed answer! I’m thinking where to put my system prompt with strict instructions on dialogue stages and the criteria on when AI can move on.
Also if I got you right it’s better for me to put the role in “model reminder” section as well.
@danny_support can you unveil details on how exactly those field work. When every of them supposed to be sent to an endpoint?
The Role prompt is visible to the model throughout the entire conversation and does not degrade over time, so you can use it for your core instructions or rules that need to be consistently enforced. The Model Reminder is the very first instruction the model sees for each user message, so it gets the most emphasis and is ideal for special guidance or critical reminders you want to reinforce on every interaction.
If you need strict control over dialogue stages or criteria, keep your main system prompt in the Role field. Use the Model Reminder for anything that needs extra emphasis or for important rules that should always be front and center as the conversation continues.
Which means both of those fields are presented to a model in every user messages, correct?
The Role prompt is like a banner on the wall that the AI sees the whole time, so put your “always follow these rules” instructions there. The Model Reminder is a fresh sticky note placed on top before every single reply, giving extra weight to whatever you write on it. Use Role for core guidance such as tone, length, or don’ts, and use Model Reminder for urgent or time-sensitive tips you want highlighted every turn.
I can understand that it can be challenging to understand why both things exist, but they serve a purpose. I will explain it using another simple example.
Think of the Role prompt as the handbook that always sits open on the desk, setting the permanent rules, while the Model Reminder is the supervisor who leans in before every single reply and whispers a quick, must-follow instruction for that moment.
I hope this helps ![]()
Hey @conrad here’s a deep dive based on my experience building on Pickaxe for over 20 clients over the past 8 months:
It’s a crucial one for building robust and reliable AI tools. Your understanding is correct, and we can clarify it further. The best place for your strict instructions depends entirely on whether you are building a
Chatbot or a Form .
Let’s break it down.
1. Role vs. Model Reminder
First, let’s solidify the difference between the two fields, as your initial explanation was perfect:
- Role Section: This is for your primary system prompt. It defines the AI’s core purpose, persona, and what it is supposed to do. Think of it as the tool’s permanent Standard Operating Procedure (SOP).
- Model Reminder (in Configure tab): This is a supplementary instruction aimed directly at the underlying Large Language Model (e.g., GPT-4o). It’s not for defining the agent’s task but for providing high-level context to override the model’s default “factory settings.” Your electrician example is the canonical use case - informing the model that its users are professionals and it should therefore refrain from giving basic, unhelpful safety warnings.
2. Where to Put Your Strict Instructions
Now, for your main question about dialogue stages and criteria. Here is the best practice:
If you are building a CHATBOT:
For a conversational chatbot, all instructions - the role, persona, rules, and dialogue stages - should go into the single
Role prompt box in the Prompt tab. A great way to keep it organized is by using Markdown headings.
Example Structure:
# Role
You are a career coach helping users prepare for job interviews. Your goal is to run the user through a mock interview and provide feedback.
## Workflow Stages
1. **Greeting & Role Selection:** Greet the user and ask them what role they are interviewing for.
2. **Mock Interview:** Once they provide the role, begin asking them common interview questions for that position.
3. **Feedback:** After 5 questions, stop the interview and provide constructive feedback on their answers.
4. **Closing:** Ask if they would like to practice more or end the session.
## Rules
- Never ask more than 5 questions before providing feedback.
- Only move to the Feedback stage after the user has answered the 5th question.
- Always remain encouraging and supportive.
If you are building a FORM:
For a form-based tool, the instructions are split into two separate locations. This is a critical difference.
- Role Box: This should only contain the AI’s persona.
- Example:
You are an expert copywriter specializing in LinkedIn posts.
- Prompt Frame: This is where you put all the strict instructions, rules, and logic for how the AI should process the user’s inputs from the form fields.
- Example:
Analyze the user's {topic_input} and {tone_selection}. Write a 300-word LinkedIn post. Start with a viral hook and end with a clear call to action.
Final Clarification
To address your last point directly: do not put the Role instructions into the Model Reminder section. This would be redundant and could confuse the AI. The Model Reminder has a very specific and separate purpose.
Quick Rule of Thumb:
- Role Section: The “Who” and “What” of your AI.
- Strict Instructions/Workflow: The “How” of the task. (Place this after the Role in a Chatbot’s prompt, or in the Prompt Frame for a Form).
- Model Reminder: A special note to the underlying model to adjust its default behavior (e.g., “The user is a certified professional, do not give them beginner advice”).
By structuring your prompts this way, you make them cleaner, easier to debug, and more reliable.
Hope this helps!