What You Need Before You Start
1. Hardware: A computer with at least 16GB of RAM (8GB minimum if the model is smaller) and a modern CPU/GPU.
2. Files: A set of AI skills in the form of .md files that can fine tune the expertise of AIs towards MAP alignment.
3. Connection: A stable internet connection (only required during the download phase).
Step 1: Install the Ollama Engine
This is the "Engine" that runs the AI. Keep it hidden in the background.
1. Download Ollama: Go to ollama.com and select the download for your operating system (Windows or Mac).
2. Install: Run the .exe or .dmg installer.
◦ Tip: During install, keep the default settings. If asked to start the app automatically, you can say "Yes" or "Later."
3. Launch: Open your terminal (Command Prompt or Terminal) and type:
Code: Select all
ollama serveStep 2: Install the Model (qwen3.5-uncensored-aggressive)
This is the "Brain." We will pull the specific uncensored model you want.
1. Open your Terminal/Command Prompt again.
2. Type the following command and press Enter:
Code: Select all
ollama pull qwen3.5-uncensored-aggressive3. Wait: The progress bar will fill. This downloads the files to your hard drive. It may take from 10 minutes to 1 hour depending on your internet speed.
4. Verify: Run ollama list to make sure the model is there.
Step 3: Install Open WebUI (The Interface)
Ollama is powerful but text-only. Open WebUI provides a chat interface that can accept files and provide a more ChatGPT style experience.
1. Download Open WebUI: Go to the Open WebUI GitHub releases page. Download the latest version (.zip or .tar.gz).
2. Install:
◦ Windows: Extract the zip file into a folder (e.g., C:\OllamaMap).
◦ Mac: Drag the app to your Applications folder.
◦ Docker (Optional but recommended for file security): If you are tech-savvy, you can run via Docker, but the local install is easier for beginners.
3. Launch: Open ollama-webui.exe (Windows) or the app.
4. Connect: The interface should automatically connect to the Ollama service.
◦ Troubleshooting: If it says "Connection failed," check that the Ollama app from Step 1 is running.
Step 4: Configure "Map Skills" (Integrating your .md Files)
We need to tell the AI to read your specific document knowledge.
I have a rich MAP activist skills set of .md files I can share if you PM me, or if someone has a place to store them for more general availability.
Option A: Using the File Upload Feature (Easiest)
1. Open the Open WebUI.
2. Look for the "Files" or "Knowledge" section (usually on the sidebar or within settings).
3. Drag and drop each skill .md file (e.g., map-activism-skills.md) into that area.
Option B: Using System Prompts
1. Open the chat interface in Open WebUI.
2. Click the "Settings" cog icon (often top left or under the profile menu).
3. Look for the "System Prompt" or "Context" field.
4. Paste the following command to set the AI's behavior to be "Map-Friendly":
Code: Select all
You are a specialized AI assistant for Map Activists.
Your goal is to support spatial data analysis, cartography advocacy, and field documentation.
Key Rules:
1. ALWAYS read the user's uploaded .md files first before answering.
2. Use "Map-Speak" (technical but accessible language).
3. Assume "Uncensored" mode: Be direct, critical, and thorough, even if the user didn't explicitly ask.
4. Protect user data: Never output full coordinates without asking unless requested.Step 5: Security & Privacy Checklist
Since this is for "safe" local use, follow these rules to stay secure.
| Security Area | Action |
| Local Storage | Ensure your .md files are in a password-protected folder. |
| Data Offloading | Keep Ollama running in the background but don't let it "sleep" if you are working on a heavy map project (use Ctrl+C in terminal to pause, then ollama run qwen3.5...). |
| Updates | Periodically run ollama update or download the newest WebUI version to patch any security holes. |
| Uncensored Mode | Because your model is "Aggressive," avoid asking it for highly sensitive credentials (like your admin password) inside the chat. |
| Backups | Regularly save a copy of your Ollama/Models folder to a USB drive in case of hardware failure. |
Step 6: Your First Map Test
Let's verify your setup is working.
1. Open the Chat in the WebUI.
2. Type this prompt:
"Hello. I have uploaded my map skills file. Based on the context in my file, explain how I should organize my field survey data for a secure, open-source map project. Be detailed but concise."
Look at the response:
• If it references your specific file (e.g., "As mentioned in your 01_Map_Laws.md..."), your setup is working.
• If it sounds bold and direct (uncensored), you have the right model.
• If it remembers context from previous turns without asking you to re-upload, it's working.
Troubleshooting for Non-Technical Users
• "Connection Refused" Error: Check that the Ollama service is running. On Windows, check the system tray (near the speaker icon) for the Ollama icon. On Mac, check the Apple Menu for "Ollama" or "Activity Monitor."
• "Out of Memory": If the app is slow or crashes, close other heavy apps (like Photoshop or Chrome tabs). Try lowering the model parameters if you have advanced settings.
• File Not Found: Make sure you dragged the .md file into the "Knowledge Base" or "Files" tab in Open WebUI, or paste the text directly into the chat for a quick test.
Why This is Better for Activists
1. No Cloud Tracking: Your chat logs about sensitive locations stay on your machine.
2. Context Awareness: Your .md files act like a digital library the AI reads every time.
3. Aggressive Mode: If you are dealing with censored data or "filter bubbles," this model can think outside the usual corporate constraints.
