Documentation Index

Fetch the complete documentation index at: https://docs.unqork.io/llms.txt

Use this file to discover all available pages before exploring further.

Understanding AI Model Settings

Prev Next

UnqorkAI Beta components like the AI Agent component and AI Summarizer component expose settings that control how the underlying Gemini model generates responses. These settings are familiar to AI practitioners but can feel opaque to Creators who are new to working with large language models.

This guide explains each setting so you can make informed decisions when configuring AI components.

System Prompt

The system prompt is a set of instructions you provide to the AI model before the conversation or summarization begins. The model uses the system prompt to understand its role, tone, and constraints.

Think of the system prompt as a briefing you give to an expert before they begin a task. A well-written system prompt produces more consistent, relevant, and appropriate responses.

   

Note: The AI Summarizer component calls this setting Instructions instead of System Prompt. The behavior is the same.

   

The AI Summarizer component calls this setting Instructions instead of System Prompt. The behavior is the same.

Example system prompts:

   

You are a customer support specialist for a retail brand. Answer questions politely and concisely. If you do not know the answer, say so and suggest the customer contact support.

   

You are a document summarizer. Summarize the provided text in three bullet points. Focus on key decisions, action items, and deadlines.

Tips for writing effective system prompts:

       
  •        

    Be specific about the agent's role and tone.

       
  •    
  •        

    Define what the agent should do when it does not know an answer.

       
  •    
  •        

    If output format matters, specify it in the system prompt using bullet points, paragraph, numbered list, and so on.

       

Temperature

Temperature controls how creative or predictable the model's responses are, on a scale from 0 to 2. Lower values produce more predictable responses; higher values produce more creative ones.

                                                                                                                                                                                                                                                                                            
                   

Value

               
                   

Effect

               
                   

Low (0–0.3)

               
                   

Responses are more focused, consistent, and predictable. Use for factual Q&A, document retrieval, or any context where accuracy matters more than variety.

               
                   

Medium (0.4–0.8)

               
                   

Balanced responses. The model is helpful and natural without being unpredictable. Good for most use cases.

               
                   

High (0.9–2)

               
                   

Responses are more varied and creative. Use for brainstorming, creative writing, or scenarios where novelty is valuable.

               

The AI Agent defaults to 0.5. For most customer-facing applications, a lower temperature produces more reliable results.

Thinking Level

Thinking Level is a simplified control for reasoning depth, available in the AI Summarizer component. It replaces the need to set the Thinking Token Budget directly for most use cases.

                                                                                                                                                                                                                                                                                                                                                    
                   

Level

               
                   

Effect

               
                   

Off

               
                   

No internal reasoning. Fastest responses.

               
                   

Low

               
                   

Minimal reasoning. Suitable for straightforward summarization tasks.

               
                   

Medium

               
                   

Moderate reasoning. Good for longer documents or content that requires nuance.

               
                   

High

               
                   

Extended reasoning. Use for complex content where accuracy is critical.

               

The AI Summarizer defaults to Low. For fine-grained control over reasoning depth, use the Thinking Token Budget setting instead.

Thinking Token Budget

The Thinking Token Budget controls how much internal reasoning the model performs before generating a response. Higher values give the model more capacity to reason through complex questions before replying.

This setting is relevant for models that support extended thinking, a mode where the model works through a problem step by step internally before producing output.

                                                                                                                                                                                                                                                                                            
                   

Value

               
                   

Effect

               
                   

Low (128–512)

               
                   

Minimal internal reasoning. Fast responses, suitable for straightforward Q&A.

               
                   

Medium (512–4096)

               
                   

Moderate reasoning. Good for questions that require the model to consider multiple factors.

               
                   

High (4096+)

               
                   

Extended reasoning. Use for complex multi-step problems, document analysis, or scenarios where accuracy is critical.

               
   

Note: The AI Agent defaults to 128. Increasing the Thinking Token Budget increases processing time and cost.

Max Output Tokens

Max Output Tokens sets the maximum length of the model's response. One token is roughly equivalent to four characters of text.

                                                                                                                                                                                                                                                                                                                                                    
                   

Value

               
                   

Approximate output

               
                   

256

               
                   

A short paragraph

               
                   

1024

               
                   

Several paragraphs

               
                   

2048

               
                   

A detailed multi-paragraph response

               
                   

4096

               
                   

A long, comprehensive response

               

The AI Agent defaults to 2048. If responses are being cut off mid-sentence, increase this value. If responses are unnecessarily long, decrease it.

   

Note: Max Output Tokens sets a ceiling, not a target. The model generates as much text as needed up to the limit. It does not pad responses to fill the token budget.

Choosing the Correct Settings

Most use cases work well with the default settings. Adjust when you notice the following issues:

                                                                                                                                                                                                                                                                                                                                                                                                                                                                    
                   

Issue

               
                   

Adjustment

               
                   

Responses feel generic or repetitive

               
                   

Increase Temperature

               
                   

Responses are inconsistent or off-topic

               
                   

Decrease Temperature

               
                   

The agent struggles with complex questions

               
                   

Increase Thinking Token Budget

               
                   

Summaries lack depth or miss nuance

               
                   

Increase Thinking Level

               
                   

Responses are cut off before finishing

               
                   

Increase Max Output Tokens

               
                   

Responses are unnecessarily long

               
                   

Decrease Max Output Tokens

               

   
                   

Changelog

       
       
           
                                                                                                                                                                                                                                                                                                                                                                            
                               

Date

                           
                               

Change

                           
                               

                           
                               

Initial publication.