Today we are building a release note generator that turns raw git commits into polished Markdown summaries. I use this to draft weekly changelogs without staring at a blank page. Because Oxlo.ai uses flat per-request pricing, loading an entire week of verbose commit history does not inflate the cost.
What you'll need
- Python 3.10 or newer
- The OpenAI SDK:
pip install openai - An Oxlo.ai API key from https://portal.oxlo.ai
Step 1: Instantiate the Oxlo.ai client
First, I import the OpenAI SDK and point it at Oxlo.ai. The base URL and key are the only differences from the standard setup.
from openai import OpenAI
client = OpenAI(base_url="https://api.oxlo.ai/v1", api_key="YOUR_OXLO_API_KEY")
response = client.chat.completions.create(
model="llama-3.3-70b",
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Say 'Connection OK'"},
],
)
print(response.choices[0].message.content)Step 2: Define the system prompt
The system prompt is the contract. I keep it in a dedicated constant so I can iterate without touching the rest of the code.
SYSTEM_PROMPT = """You are a technical writer drafting release notes.
Rules:
- Group commits into Added, Fixed, and Changed sections.
- Write in Markdown with H2 headers.
- Ignore merge commits and minor dependency bumps.
- Keep each bullet under 120 characters.
- Do not mention commit hashes."""Step 3: Feed raw commits into the context window
I format the raw data as a plain text list. Because Oxlo.ai bills per request rather than per token, I can pass the full week of commits in one shot without worrying about context length.
from openai import OpenAI
client = OpenAI(base_url="https://api.oxlo.ai/v1", api_key="YOUR_OXLO_API_KEY")
commits = [
{"hash": "a1b2c3d", "message": "feat: add OAuth2 login flow", "author": "dev1"},
{"hash": "e5f6g7h", "message": "fix: resolve race condition in worker queue", "author": "dev2"},
{"hash": "i8j9k0l", "message": "chore: bump pytest to 8.2", "author": "dev3"},
]
user_message = "Here are the commits since last release:\n"
for c in commits:
user_message += f"- {c['message']} ({c['author']})\n"
response = client.chat.completions.create(
model="llama-3.3-70b",
messages=[
{"role": "system", "content": SYSTEM_PROMPT},
{"role": "user", "content": user_message},
],
)
print(response.choices[0].message.content)Step 4: Lock the output shape with JSON mode
To make the result machine readable, I force valid JSON by setting response_format. This lets me pipe the output into other tools or a static site generator.
from openai import OpenAI
import json
client = OpenAI(base_url="https://api.oxlo.ai/v1", api_key="YOUR_OXLO_API_KEY")
SYSTEM_PROMPT = """You are a technical writer drafting release notes.
Return strictly valid JSON with this shape:
{
"added": ["string"],
"fixed": ["string"],
"changed": ["string"]
}
Rules:
- Group items logically.
- Ignore merge commits and minor dependency bumps.
- Keep each bullet under 120 characters."""
commits = [
{"hash": "a1b2c3d", "message": "feat: add OAuth2 login flow", "author": "dev1"},
{"hash": "e5f6g7h", "message": "fix: resolve race condition in worker queue", "author": "dev2"},
{"hash": "i8j9k0l", "message": "chore: bump pytest to 8.2", "author": "dev3"},
]
user_message = "Here are the commits since last release:\n"
for c in commits:
user_message += f"- {c['message']} ({c['author']})\n"
response = client.chat.completions.create(
model="llama-3.3-70b",
messages=[
{"role": "system", "content": SYSTEM_PROMPT},
{"role": "user", "content": user_message},
],
response_format={"type": "json_object"},
)
result = json.loads(response.choices[0].message.content)
print(json.dumps(result, indent=2))Step 5: Package the generator into a CLI script
I wrap the logic in a small script that reads commit lines from stdin. This is the version I drop into CI pipelines.
import sys
import json
from openai import OpenAI
client = OpenAI(base_url="https://api.oxlo.ai/v1", api_key="YOUR_OXLO_API_KEY")
SYSTEM_PROMPT = """You are a technical writer drafting release notes.
Return strictly valid JSON with this shape:
{
"added": ["string"],
"fixed": ["string"],
"changed": ["string"]
}
Rules:
- Group items logically.
- Ignore merge commits and minor dependency bumps.
- Keep each bullet under 120 characters."""
def generate_notes(commit_lines, model="llama-3.3-70b"):
user_message = "Here are the commits since last release:\n" + "".join(commit_lines)
response = client.chat.completions.create(
model=model,
messages=[
{"role": "system", "content": SYSTEM_PROMPT},
{"role": "user", "content": user_message},
],
response_format={"type": "json_object"},
)
return json.loads(response.choices[0].message.content)
if __name__ == "__main__":
raw = sys.stdin.readlines()
notes = generate_notes(raw)
print(json.dumps(notes, indent=2))Run it
Save the file as notes.py, create a small input file, and pipe it through.
$ printf "feat: add OAuth2 login flow (dev1)\nfix: resolve race condition in worker queue (dev2)\nchore: bump pytest to 8.2 (dev3)\nfeat: introduce caching layer for configs (dev1)\nfix: patch SQL injection in search endpoint (dev2)\n" | python notes.pyExample output:
{
"added": [
"OAuth2 login flow for secure third-party integrations.",
"In-memory caching layer for hot configuration paths."
],
"fixed": [
"Race condition in worker queue that caused duplicate jobs.",
"SQL injection vulnerability in the search endpoint."
],
"changed": []
}Next steps
Pipe the JSON output directly into a Jinja template to generate static HTML for your docs site. If you need deeper reasoning on security-related commits, route only those items through deepseek-v3.2 for a detailed impact paragraph.
Top comments (0)