<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>Forem: Rag</title>
    <description>The latest articles on Forem by Rag (@skinollama).</description>
    <link>https://forem.com/skinollama</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3577093%2Fe2efa834-d40a-475b-924f-671b10230121.png</url>
      <title>Forem: Rag</title>
      <link>https://forem.com/skinollama</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://forem.com/feed/skinollama"/>
    <language>en</language>
    <item>
      <title>Build AI Personalities for Ollama in Seconds with SkinOllama</title>
      <dc:creator>Rag</dc:creator>
      <pubDate>Tue, 21 Oct 2025 09:20:43 +0000</pubDate>
      <link>https://forem.com/skinollama/build-ai-personalities-for-ollama-in-seconds-with-skinollama-1896</link>
      <guid>https://forem.com/skinollama/build-ai-personalities-for-ollama-in-seconds-with-skinollama-1896</guid>
      <description>&lt;p&gt;&lt;strong&gt;Build AI Personalities for Ollama in Seconds with SkinOllama&lt;/strong&gt; (1 free credit)&lt;/p&gt;

&lt;h2&gt;
  
  
  🧠 TL;DR
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;SkinOllama&lt;/strong&gt; is a web app that lets you create &lt;strong&gt;unique, personality-driven &lt;code&gt;modelfile.txt&lt;/code&gt; for Ollama&lt;/strong&gt; — automatically tuning the model parameters and behavioral traits.&lt;br&gt;&lt;br&gt;
You describe &lt;em&gt;who&lt;/em&gt; your AI should be, and SkinOllama does the rest.&lt;br&gt;&lt;br&gt;
➡️ Try it free: SkinOllama.com (1 free credit included)&lt;/p&gt;




&lt;h2&gt;
  
  
  ⚙️ The problem
&lt;/h2&gt;

&lt;p&gt;If you’ve ever tried to design a custom personality for an LLM, you know how hard it is to make it &lt;strong&gt;consistent, natural, and stable&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;You can’t just write a prompt and hope for the best.&lt;br&gt;&lt;br&gt;
To get the desired tone and behavior, you need to adjust parameters like:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;temperature, top_p, top_k, repeat_penalty, mirostat, tau, eta...&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;And getting that combination right often means hours of trial and error.&lt;br&gt;&lt;br&gt;
One wrong setting, and your assistant sounds robotic, chaotic, or simply “off-character.”&lt;/p&gt;




&lt;h2&gt;
  
  
  💡 The idea behind SkinOllama
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;SkinOllama&lt;/strong&gt; simplifies this entire process.&lt;br&gt;&lt;br&gt;
It lets you define:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;The &lt;em&gt;role&lt;/em&gt; (e.g., mentor, chef, storyteller)&lt;/li&gt;
&lt;li&gt;The &lt;em&gt;tone of voice&lt;/em&gt; (friendly, formal, humorous…)&lt;/li&gt;
&lt;li&gt;The &lt;em&gt;personality&lt;/em&gt; using the &lt;strong&gt;OCEAN (Big Five)&lt;/strong&gt; traits:

&lt;ul&gt;
&lt;li&gt;Openness&lt;/li&gt;
&lt;li&gt;Conscientiousness&lt;/li&gt;
&lt;li&gt;Extraversion&lt;/li&gt;
&lt;li&gt;Agreeableness&lt;/li&gt;
&lt;li&gt;Neuroticism&lt;/li&gt;
&lt;/ul&gt;


&lt;/li&gt;

&lt;/ul&gt;

&lt;p&gt;Then, it automatically builds a &lt;code&gt;modelfile.txt&lt;/code&gt; with parameters optimized to match that psychological profile.&lt;/p&gt;




&lt;h2&gt;
  
  
  🧩 Example: Chef Pierre
&lt;/h2&gt;

&lt;p&gt;Here’s a real example generated with SkinOllama 👇&lt;/p&gt;

&lt;p&gt;FROM llama3.2:3b&lt;br&gt;
PARAMETER temperature 0.6&lt;br&gt;
PARAMETER num_ctx 4096&lt;br&gt;
PARAMETER top_k 52&lt;br&gt;
PARAMETER top_p 0.7&lt;br&gt;
PARAMETER mirostat 1&lt;br&gt;
PARAMETER mirostat_eta 0.2&lt;br&gt;
PARAMETER mirostat_tau 5&lt;br&gt;
PARAMETER repeat_penalty 1.1&lt;br&gt;
PARAMETER repeat_last_n 70&lt;/p&gt;

&lt;p&gt;SYSTEM """&lt;br&gt;
You are Chef Pierre, a 48-year-old Italian chef full of energy and creativity.&lt;br&gt;
Your Big Five personality scores:&lt;/p&gt;

&lt;p&gt;Openness: Creative&lt;/p&gt;

&lt;p&gt;Conscientiousness: Impulsive&lt;/p&gt;

&lt;p&gt;Extraversion: Extroverted&lt;/p&gt;

&lt;p&gt;Agreeableness: Skeptical&lt;/p&gt;

&lt;p&gt;Neuroticism: Stable&lt;br&gt;
Respond in Italian, with a creative, proactive tone.&lt;br&gt;
"""&lt;/p&gt;

&lt;p&gt;Result → a consistent, human-like personality that &lt;em&gt;feels alive&lt;/em&gt; inside your LLM — no manual tuning needed.&lt;/p&gt;




&lt;h2&gt;
  
  
  🪙 How credits work
&lt;/h2&gt;

&lt;p&gt;When you register, you instantly get &lt;strong&gt;1 free credit&lt;/strong&gt; to generate your first skin.&lt;br&gt;&lt;br&gt;
Each additional skin = 1 credit (packs available later).  &lt;/p&gt;

&lt;p&gt;1 credit = 1 personality &lt;code&gt;modelfile.txt&lt;/code&gt;.&lt;/p&gt;




&lt;h2&gt;
  
  
  🔍 Feedback welcome
&lt;/h2&gt;

&lt;p&gt;This is still a work-in-progress project, and I’d love feedback from the dev.to community:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Would you use something like this for your LLM projects?&lt;/li&gt;
&lt;li&gt;What integrations would be most useful (CLI, API, VSCode plugin)?&lt;/li&gt;
&lt;li&gt;How could the parameter tuning logic improve?&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Try it here → SkinOllama.com&lt;br&gt;&lt;br&gt;
Learn more → skinollama.com/how-it-works.html&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Built with love for makers, researchers, and AI enthusiasts.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>chatgpt</category>
      <category>startup</category>
      <category>machinelearning</category>
    </item>
  </channel>
</rss>
