Welcome to the comprehensive documentation for Minibase Ollama, your tool for running fine-tuned AI models locally with full privacy and control.
What is Minibase Ollama?
Minibase Ollama is a specialized version of Ollama that seamlessly integrates with your Minibase account to run your custom-trained models locally. It maintains full compatibility with the standard Ollama ecosystem while adding powerful features specific to your Minibase models.
Key Features
Secure Authentication - Automatic API key integration with your Minibase account
Smart Instructions - Your models come pre-configured with their training instructions
Custom Override - Use `--instruction` flag to customize model behavior on-the-fly
Full Ollama Compatibility - Works with all Ollama tools, libraries, and integrations
Local Execution - Your data never leaves your machine
Easy Updates - Simple model management and version control
Documentation Guides
Getting Started
Installation and Setup Guide:
Download and install Minibase Ollama for your platform
Configure your API key and registry
Verify installation
System requirements
Quick Start Tutorial:
Download your first model
Run your first inference
Understand the basics
Common first-time workflows
Core Usage
CLI Command Reference:
Complete list of all commands
All available flags and options
Command examples
Exit codes and error handling
Working with Instructions:
Understanding default instructions
Using custom instructions with `--instruction`
Best practices for instruction design
Examples for different use cases
Model Management Guide:
Pulling models from Minibase
Listing available models
Deleting models
Model naming conventions
Storage locations
Integration & Development
HTTP API Integration:
Direct API calls with curl/HTTP clients
API endpoints reference
Request/response formats
Streaming responses
Error handling
Python SDK Integration:
Using the official Ollama Python library
Integration patterns
Code examples for common tasks
Best practices
Advanced Usage Guide:
Response formatting (JSON, etc.)
Streaming vs non-streaming
Model options and parameters
Keep-alive and performance tuning
Multi-turn conversations
Support
Troubleshooting & FAQ:
Common issues and solutions
Frequently asked questions
Performance optimization
Getting help
Quick Links
External Resources
Minibase-Specific Features
The key differences between standard Ollama and Minibase Ollama:
Feature | Standard Ollama | Minibase Ollama |
Model Source | Public registry (ollama.com) | Your Minibase account |
Authentication | None (public models) | API key authentication |
Model Instructions | Manual via Modelfile | Automatic from training |
Custom Instructions | Via Modelfile only | `--instruction` flag |
Model Format | Community models | Your fine-tuned models |
Common Use Cases
1. Content Moderation
```bash
# Your model automatically knows to remove toxicity
echo "Your toxic text here" | ollama run my-toxic-model
```
2. Text Classification
```bash
# Classify text with your custom-trained classifier
echo "Email content here" | ollama run my-spam-classifier
```
3. Custom Translation
```bash
# Override default instruction for translation
echo "Hello world" | ollama run my-model --instruction "Translate to Spanish"
```
4. API Integration
```python
import ollama
response = ollama.chat(model='my-model', messages=[
{'role': 'user', 'content': 'Process this text'}
])
print(response['message']['content'])
```
5. Batch Processing
```bash
# Process multiple files
for file in *.txt; do
cat "$file" | ollama run my-model > "processed_$file"
done
```
System Architecture
Minibase Ollama consists of three main components:
Your Application - CLI, Python, JavaScript, curl, or any HTTP client
Ollama Server (localhost:11434) - Handles model loading, inference, memory management, and GPU acceleration
Local Model Storage (~/.ollama/models/) - Stores model files (GGUF format), manifests with instructions, and configuration
On first use, models are downloaded from Minibase. Subsequent uses access cached local files for fast inference.
Quick Command Reference
```bash
# Download a model
ollama pull my-model-id
# Run inference (uses default instruction)
echo "input text" | ollama run my-model-id
# Run with custom instruction
echo "input text" | ollama run my-model-id --instruction "Custom task"
# List downloaded models
ollama list
# View model details
ollama show my-model-id
# Delete a model
ollama rm my-model-id
# Start server manually (usually auto-starts)
ollama serve
```
Support and Community
Getting Help
Check the Documentation - Start with the guides above
Review Troubleshooting - See Troubleshooting Guide
Contact Support - Email [email protected]
Platform Issues - Check Minibase Help Center
Reporting Issues
When reporting issues, please include:
Operating system and version
Minibase Ollama version (`ollama --version`)
Model ID you're trying to use
Complete error message
Steps to reproduce
What's Next?
Ready to get started? Follow these steps:
Download and Install - Get Minibase Ollama running
Quick Start - Run your first model
Explore Features - Learn all the commands
Integrate - Add to your applications