How AI Can Remember You: Simple Setup Guide

Yesterday morning, I sat at my desk in Munich and opened Ollama to start another day. The model greeted me as it always does, but something felt off. It was like talking to a stranger who had no idea about our ongoing projects or the context of our work together. I realized then that while the AI could answer questions efficiently, it lacked the crucial memory that would make it a true assistant.

The difference between a chatbot and an assistant lies in their ability to remember context and recall previous interactions. A model with memory understands your style, knows your projects, and can provide more accurate and relevant answers without you having to repeat yourself every time. To turn my local AI into an assistant, I needed to give it some form of long-term memory.

The first step is creating a system prompt, a short text file that contains essential information about me and my work. This prompt serves as the model's foundation for understanding who it's assisting and what kind of support is needed. For example, I created a simple text file named `system-prompt.txt` with just four sentences:

You are an assistant for Max, who runs a small premium fashion brand called HACOY, based in Munich. Max also runs a personal blog, paints in his free time, and lives with ADHD. He likes calm, direct answers. He does not like jargon.

Every time I start a new conversation with the AI, I paste this file as the first message. This ensures that the model is always aware of my context and can tailor its responses accordingly. The system prompt acts like a mental reset for the AI, allowing it to remember key details about me without needing to be reminded repeatedly.

The second level of memory involves saving and reusing conversation history. When I finish an important discussion with the AI, I save the entire chat as a text file in a folder labeled `conversations`. At the beginning of my next session, I simply paste this saved conversation into the new chat window. This way, the AI picks up where we left off, remembering previous questions and answers, making subsequent interactions smoother and more efficient.

Using these two methods, system prompts and saved conversation history, I transformed Ollama from a generic chatbot into an assistant that truly understands my needs and work style. The improvement in response quality is significant. Instead of generic replies, the AI now provides personalized advice based on our ongoing projects and past conversations.

Setting up these features may seem like a small change, but it has had a profound impact on how I interact with my local AI. It feels as though I have an intelligent assistant that knows me well, rather than just another tool that requires constant reorientation. This level of personalization is invaluable for founders and small teams who rely on quick, context-aware support to navigate their daily tasks.

Remember, the key to effective memory in your AI model lies in simplicity and relevance. A good system prompt should be concise yet comprehensive, covering the essentials without overwhelming the model with unnecessary details. Similarly, saving conversation history helps maintain continuity and ensures that every interaction builds upon previous ones.

In the end, giving your local AI a sense of memory is like equipping it with an understanding brain rather than just a reactive mind. It's not about making the model smarter but about making it more aware of its surroundings, your projects, your preferences, and your context. This small tweak can transform how you work with your AI, turning it from a passive tool into a proactive assistant that truly understands and supports you.