Skip to content
kit

Using Ollama with kit review

🦙 Using Ollama with Kit

Kit has first-class support for free local AI models via Ollama. No API keys, no costs, no data leaving your machine.

Why Ollama?

  • No cost - unlimited usage
  • Complete privacy - data never leaves your machine
  • No API keys - just install and run
  • No rate limits - only hardware constraints
  • Works offline - perfect for secure environments
  • Latest models - access to cutting-edge open source AI

Quick Setup (2 minutes)

1. Install Ollama

Terminal window
# macOS/Linux
curl -fsSL https://ollama.ai/install.sh | sh


# Windows
# Download from https://ollama.ai/download

2. Pull a Model

Choose based on your use case:

Terminal window
# Best for code tasks (recommended)
ollama pull qwen2.5-coder:latest


# Best for reasoning
ollama pull deepseek-r1:latest


# Best for coding agents
ollama pull devstral:latest


# Good general purpose
ollama pull llama3.3:latest

3. Start Using with Kit

from kit import Repository
from kit.summaries import OllamaConfig


# Configure Ollama
config = OllamaConfig(model="qwen2.5-coder:latest")


# Use with any repository
repo = Repository("/path/to/your/project")
summarizer = repo.get_summarizer(config=config)


# Summarize code at no cost
summary = summarizer.summarize_file("main.py")
print(summary)  # Cost: $0.00

💡 Complete Examples

Code Summarization

from kit import Repository
from kit.summaries import OllamaConfig


# Setup
repo = Repository("/path/to/project")
config = OllamaConfig(
    model="qwen2.5-coder:latest",
    temperature=0.1,  # Lower for more focused analysis
    max_tokens=1000
)
summarizer = repo.get_summarizer(config=config)


# Summarize a file
summary = summarizer.summarize_file("complex_module.py")
print(f"File Summary: {summary}")


# Summarize a specific function
func_summary = summarizer.summarize_function("utils.py", "parse_config")
print(f"Function Summary: {func_summary}")


# Summarize a class
class_summary = summarizer.summarize_class("models.py", "UserManager")
print(f"Class Summary: {class_summary}")

PR Reviews (No Cost)

Terminal window
# Setup Ollama for PR reviews
kit review --init-config
# Choose "ollama" as provider
# Choose "qwen2.5-coder:latest" as model


# Review any PR at no cost
kit review https://github.com/owner/repo/pull/123
# Cost: $0.00

Batch Documentation Generation

from kit import Repository
from kit.summaries import OllamaConfig
import os


def generate_docs(project_path, output_file):
    """Generate documentation for an entire project."""
    repo = Repository(project_path)
    config = OllamaConfig(model="qwen2.5-coder:latest")
    summarizer = repo.get_summarizer(config=config)


    with open(output_file, 'w') as f:
        f.write(f"# Documentation for {os.path.basename(project_path)}\n\n")


        # Get all Python files
        files = repo.get_file_tree()
        python_files = [f for f in files if f['path'].endswith('.py') and not f.get('is_dir')]


        for file_info in python_files:
            file_path = file_info['path']
            try:
                summary = summarizer.summarize_file(file_path)
                f.write(f"## {file_path}\n\n{summary}\n\n")
                print(f"✅ Documented {file_path} (Cost: $0.00)")
            except Exception as e:
                print(f"⚠️ Skipped {file_path}: {e}")


# Usage
generate_docs("/path/to/project", "project_docs.md")

Legacy Codebase Analysis

def analyze_legacy_code(repo_path):
    """Analyze and understand legacy code using free AI."""
    repo = Repository(repo_path)
    config = OllamaConfig(model="qwen2.5-coder:latest")
    summarizer = repo.get_summarizer(config=config)


    # Find all symbols
    symbols = repo.extract_symbols()


    # Group by type
    functions = [s for s in symbols if s.get('type') == 'FUNCTION']
    classes = [s for s in symbols if s.get('type') == 'CLASS']


    print(f"Found {len(functions)} functions and {len(classes)} classes")


    # Analyze complex functions (those with many lines)
    complex_functions = [f for f in functions if len(f.get('code', '').split('\n')) > 20]


    for func in complex_functions[:5]:  # Analyze top 5 complex functions
        file_path = func['file']
        func_name = func['name']


        analysis = summarizer.summarize_function(file_path, func_name)
        print(f"\n📝 {func_name} in {file_path}:")
        print(f"   {analysis}")
        print(f"   Cost: $0.00")

⚙️ Advanced Configuration

Custom Configuration

config = OllamaConfig(
    model="qwen2.5-coder:32b",        # Use larger model for better results
    base_url="http://localhost:11434", # Default Ollama endpoint
    temperature=0.0,                   # Deterministic output
    max_tokens=2000,                   # Longer responses
)

Remote Ollama Server

config = OllamaConfig(
    model="qwen2.5-coder:latest",
    base_url="http://your-server:11434",  # Remote Ollama instance
    temperature=0.1,
    max_tokens=1000,
)

Multiple Models for Different Tasks

# Code-focused model for functions
code_config = OllamaConfig(model="qwen2.5-coder:latest", temperature=0.1)


# Reasoning model for complex analysis
reasoning_config = OllamaConfig(model="deepseek-r1:latest", temperature=0.2)


# Use different models for different tasks
code_summarizer = repo.get_summarizer(config=code_config)
reasoning_summarizer = repo.get_summarizer(config=reasoning_config)

🔧 Troubleshooting

Common Issues

“Connection refused” error:

Terminal window
# Make sure Ollama is running
ollama serve


# Or check if it's already running
ps aux | grep ollama

“Model not found” error:

Terminal window
# Pull the model first
ollama pull qwen2.5-coder:latest


# List available models
ollama list

Slow responses:

  • Use smaller models like qwen2.5-coder:0.5b for faster responses
  • Reduce max_tokens for shorter outputs
  • Ensure sufficient RAM (8GB+ recommended for 7B models)

Performance Tips

  1. Choose the right model size:

    • 0.5B-3B: Fast, good for simple tasks
    • 7B-14B: Balanced speed and quality
    • 32B+: Best quality, requires more resources

  2. Optimize settings:

    # For speed
    config = OllamaConfig(
        model="qwen2.5-coder:0.5b",
        temperature=0.1,
        max_tokens=500
    )
    

    # For quality
    config = OllamaConfig(
        model="qwen2.5-coder:32b",
        temperature=0.0,
        max_tokens=2000
    )

  3. Hardware considerations:

    • RAM: 8GB minimum, 16GB+ recommended for larger models
    • GPU: Optional but significantly speeds up inference
    • Storage: Models range from 500MB to 400GB

💰 Cost Comparison

ProviderCost per ReviewSetup TimePrivacyOffline
Ollama$0.002 minutes✅ 100% private✅ Works offline
OpenAI GPT-4o~$0.10API key setup❌ Sent to OpenAI❌ Requires internet
Anthropic Claude~$0.08API key setup❌ Sent to Anthropic❌ Requires internet

🤝 Community