quick-start-guide

Quick Start: Git Commit Summarization with OSS Models

TL;DR: Use Ollama + Qwen2.5-Coder-1.5B for fastest setup and excellent results.

---

5-Minute Setup (Development)

Step 1: Install Ollama

# Linux/macOS
curl https://ollama.ai/install.sh | sh
 
# Or use Docker
docker run -d -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama

Step 2: Download Model

ollama pull qwen2.5-coder:1.5b

Step 3: Test

# Interactive test
ollama run qwen2.5-coder:1.5b
# Then type: "Summarize: Added user authentication with JWT tokens"

Step 4: Use via API

import requests
 
def summarize_commit(diff):
    response = requests.post('http://localhost:11434/api/generate',
        json={
            'model': 'qwen2.5-coder:1.5b',
            'prompt': f'Write a one-line summary of this git commit:\n\n{diff}',
            'stream': False,
            'options': {'temperature': 0.3}
        })
    return response.json()['response']
 
# Example usage
diff = """
diff --git a/auth.py b/auth.py
+import jwt
+def create_token(user_id):
+    return jwt.encode({'user': user_id}, SECRET_KEY)
"""
 
print(summarize_commit(diff))
# Output: "Add JWT token creation function for user authentication"

---

Alternative Models (Just Change the Pull Command)

# Smaller/Faster (0.5B)
ollama pull qwen2.5-coder:0.5b
 
# Higher Quality (3B)
ollama pull qwen2.5-coder:3b
 
# Best Quality (7B)
ollama pull qwen2.5-coder:7b
 
# Alternative: DeepSeek Coder
ollama pull deepseek-coder:1.3b
 
# Alternative: Phi-3
ollama pull phi3:mini

---

Production Setup with vLLM

Docker Deployment

docker run -d \
  --runtime nvidia --gpus all \
  -v ~/.cache/huggingface:/root/.cache/huggingface \
  -p 8000:8000 \
  --ipc=host \
  --name vllm-server \
  vllm/vllm-openai:latest \
  --model Qwen/Qwen2.5-Coder-1.5B-Instruct

Python Client (OpenAI Compatible)

from openai import OpenAI
 
client = OpenAI(
    base_url="http://localhost:8000/v1",
    api_key="not-needed"
)
 
def summarize_commit(diff):
    response = client.chat.completions.create(
        model="Qwen/Qwen2.5-Coder-1.5B-Instruct",
        messages=[{
            "role": "user",
            "content": f"Summarize this git commit in one line:\n\n{diff}"
        }],
        max_tokens=100,
        temperature=0.3
    )
    return response.choices[0].message.content

---

Batch Processing Example

import subprocess
import requests
from concurrent.futures import ThreadPoolExecutor
 
def get_recent_commits(n=10):
    """Get last N commits"""
    result = subprocess.check_output(
        ['git', 'log', f'-{n}', '--format=%H'],
        text=True
    )
    return result.strip().split('\n')
 
def get_commit_diff(commit_hash):
    """Get diff for a commit"""
    return subprocess.check_output(
        ['git', 'show', '--format=%B%n---', commit_hash],
        text=True
    )
 
def summarize_with_ollama(diff):
    """Summarize using Ollama"""
    response = requests.post('http://localhost:11434/api/generate',
        json={
            'model': 'qwen2.5-coder:1.5b',
            'prompt': f'Write a concise one-line summary:\n\n{diff}',
            'stream': False,
            'options': {'temperature': 0.3, 'num_predict': 100}
        })
    return response.json()['response']
 
def process_batch(num_commits=10):
    """Process commits in batch"""
    commits = get_recent_commits(num_commits)
 
    results = []
    for commit in commits:
        diff = get_commit_diff(commit)
        summary = summarize_with_ollama(diff)
        results.append({
            'commit': commit,
            'summary': summary
        })
        print(f"{commit[:8]}: {summary}")
 
    return results
 
# Run it
if __name__ == '__main__':
    summaries = process_batch(10)

---

Docker Compose (Full Stack)

version: '3.8'
 
services:
  ollama:
    image: ollama/ollama:latest
    ports:
      - "11434:11434"
    volumes:
      - ollama_data:/root/.ollama
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    restart: unless-stopped
 
volumes:
  ollama_data:

Usage:

docker-compose up -d
docker exec ollama ollama pull qwen2.5-coder:1.5b

---

Performance Expectations

Model	Size	Speed (t/s)	RAM/VRAM	Quality
qwen2.5-coder:0.5b	0.5B	100+	1GB	Good
qwen2.5-coder:1.5b	1.5B	70-100	2GB	Excellent
qwen2.5-coder:3b	3B	50-70	4GB	Excellent
qwen2.5-coder:7b	7B	30-50	6GB	Best

Per-commit time: <1 second Batch of 100 commits: 1-2 minutes

---

Prompt Templates

Basic

f"Summarize this git commit in one line:\n\n{diff}"

Conventional Commits

f"""Generate a conventional commit message for this diff.
Format: <type>(<scope>): <description>
Types: feat, fix, docs, style, refactor, test, chore
 
{diff}
 
Conventional commit:"""

Structured Output

f"""Analyze this commit and output JSON:
{{"type": "...", "scope": "...", "summary": "...", "files_changed": [...]}}
 
{diff}
 
JSON:"""

---

Troubleshooting

Ollama not responding

# Check if running
curl http://localhost:11434/api/tags
 
# Restart
ollama serve

Out of memory

# Use smaller model
ollama pull qwen2.5-coder:0.5b
 
# Or reduce context
OLLAMA_MAX_LOADED_MODELS=1 ollama serve

Slow inference

# Check GPU usage
nvidia-smi
 
# Pull GPU-optimized version
ollama pull qwen2.5-coder:1.5b-fp16

---

Resource Requirements Summary

Minimum (CPU only):

• RAM: 4GB
• Disk: 2GB
• Model: qwen2.5-coder:0.5b or 1.5b

Recommended (GPU):

• RAM: 8GB
• VRAM: 4GB (NVIDIA GPU)
• Disk: 5GB
• Model: qwen2.5-coder:1.5b or 3b

Production (GPU):

• RAM: 16GB
• VRAM: 6GB+
• Disk: 10GB
• Model: qwen2.5-coder:3b or 7b
• Server: vLLM

---

Next Steps

1. Install Ollama (5 minutes)
2. Pull qwen2.5-coder:1.5b (2 minutes)
3. Test with your commits (5 minutes)
4. Integrate into workflow (1 hour)
5. Deploy to production (optional)

---

Full Documentation

See /home/beengud/raibid-labs/sparky/research/git-commit-summarization-oss-models.md for comprehensive research and details.