git_changelog_writer

Git Changelog Writer

A multi-agent pipeline built with the Upsonic AI Agent Framework. This example demonstrates how to use Team with mode="sequential" to turn raw git log output into a ready-to-post Twitter/X update — two agents pass context automatically, no glue code required.

Features

• Sequential Team: Two agents run in sequence with automatic context handover (mode="sequential")
• Tech Lead Agent: Filters noise commits (chore, docs) and extracts user-facing changes from feat and fix entries
• Growth Hacker Agent: Converts the technical summary into a developer-native Twitter/X post (max 280 characters)
• No Glue Code: Output from Agent A flows into Agent B automatically — no variable passing or string parsing

Prerequisites

• Python 3.10+
• OpenAI API key

Installation

1. Navigate to this directory:

   cd examples/multi_agent/git_changelog_writer

2. Install dependencies:

   uv sync

3. Set up environment variables:

   export OPENAI_API_KEY="your-api-key"

Usage

uv run main.py

Project Structure

git_changelog_writer/
├── main.py        # Sequential Team pipeline
└── README.md      # This file

How It Works

1. Input: A changelog or git log --oneline string is passed to the first task. The example ships with mock data — swap it for real git log output in production.

2. Agent A (Tech Lead): Reads the commits, ignores housekeeping tags (chore, docs), and produces a clean technical summary of user-facing feat and fix entries.

3. Context Handover: Upsonic's mode="sequential" automatically injects Agent A's output into Agent B's task context. No manual wiring needed.

4. Agent B (Growth Hacker): Applies strict tone and formatting rules to write a single Twitter/X post (max 280 characters) that reads like an engineer sharing something useful, not a marketing announcement.

5. Output: tasks[-1].response contains the final tweet.

Example Output

Agent A (Tech Lead) summarizes:

Summary of User-Facing Changes:

1. New Feature: Telegram Interface
   - What Changed: Full Telegram bot interface for agents via webhook.
   - Why Users Care: Agents can now be used over Telegram chat.

2. Bug Fix: LLM Usage Tracking
   - What Changed: Fixed tracking for direct LLM call metrics.
   - Why Users Care: Accurate usage data in dashboards.

Agent B (Growth Hacker) turns it into a tweet:

Upsonic v0.72.0 is out.

New Telegram interface so your agents can run over chat.
CLI now starts interfaces directly — no programmatic setup.

- Telegram bot interface (webhook, messages, media)
- CLI + interface compatibility
- LLM usage tracking fix

Changelog: [link]

Notes

• No glue code: mode="sequential" handles all context passing between agents.
• Tone-tuned: Agent B's instructions use specific constraints ("no hashtags", "no 'we're thrilled'") rather than vague guidelines.
• Extendable: Add more agents to the list (e.g., Editor, Translator) — context still flows automatically.
• Cost: ~$0.01 per run (both agents combined).