# Troubleshooting Guide
Common issues and quick solutions for Minibase Ollama.
## Installation Issues
### "Command not found: ollama"
**Solution:**
```bash
# Add to PATH
export PATH="$HOME/.minibase/bin:$PATH"
echo 'export PATH="$HOME/.minibase/bin:$PATH"' >> ~/.zshrc
source ~/.zshrc
```
### "Permission denied"
**Solution:**
```bash
chmod +x ~/.minibase/bin/ollama
```
### macOS Security Warning
**Solution:**
1. System Settings → Privacy & Security
2. Click "Allow Anyway"
3. Run ollama again, click "Open"
## Connection Issues
### "Could not connect to ollama server"
**Solution:**
```bash
# Start the server
ollama serve
```
### Port Already in Use
**Solution:**
```bash
# Use different port
export OLLAMA_HOST="127.0.0.1:11435"
ollama serve
```
## Model Issues
### "Model not found"
**Causes:**
* Model ID incorrect
* Model not pulled yet
* No access to model
**Solution:**
```bash
# Verify model ID from minibase.ai
# Pull the model
ollama pull YOUR_CORRECT_MODEL_ID
# List what's downloaded
ollama list
```
### Download Fails
**Solution:**
* Check internet connection
* Verify API key: `cat ~/.minibase/config.json`
* Re-download Ollama package from minibase.ai
* Check model exists in your account
## Performance Issues
### Slow First Response
**Explanation:** Normal! First inference loads model into memory (2-5 seconds).
**Solution:**
```bash
# Keep model loaded longer
export OLLAMA_KEEP_ALIVE=1h
ollama serve
```
### Slow All Responses
**Solutions:**
* Check if GPU is being used: `ollama ps` (look for "GPU")
* Close other applications to free RAM
* Use smaller quantization model (select "low" in Minibase)
### High Memory Usage
**Solution:**
```bash
# Unload models faster
export OLLAMA_KEEP_ALIVE=5m
ollama serve
# Delete unused models
ollama rm OLD_MODEL
```
## Common Errors
### "Unknown capabilities for model"
**Solution:**
* Contact [email protected] with Model ID
* Try re-pulling: `ollama rm MODEL && ollama pull MODEL`
### "Context deadline exceeded"
**Causes:** Network timeout
**Solution:**
* Check internet connection
* Try again (downloads resume automatically)
## FAQ
### Can I use public Ollama models?
No. Minibase Ollama only works with models from your Minibase account.
### Do I need internet?
* **First time**: Yes, to download model
* **After that**: No, fully offline
### How much disk space do I need?
* Small models (135M): ~100-200 MB
* Medium models (1B): ~500 MB - 2 GB
* Large models (7B+): 4-40 GB
Check usage: `du -sh ~/.ollama/models`
### Can I use multiple models?
Yes! Download multiple models with `ollama pull`. Only one loads at a time by default.
### How do I update Minibase Ollama?
1. Download latest from minibase.ai → API Keys
2. Run installer (overwrites binary)
3. Your models and config are preserved
## Still Need Help?
1. Check [Ollama Documentation](https://github.com/ollama/ollama)
2. Email [email protected] with:
* Operating system
* `ollama --version` output
* Complete error message
* Model ID
## Diagnostic Commands
```bash
# Check version
ollama --version
# Check server
curl http://localhost:11434/api/version
# Check config (redact API key!)
cat ~/.minibase/config.json
# List models
ollama list
# Check what's loaded
ollama ps
# Check disk usage
du -sh ~/.ollama/models
```