https://github.com/sdkdeepa/agentstack
https://github.com/sdkdeepa/agentstack
Last synced: about 1 month ago
JSON representation
- Host: GitHub
- URL: https://github.com/sdkdeepa/agentstack
- Owner: sdkdeepa
- Created: 2026-06-10T07:54:45.000Z (about 1 month ago)
- Default Branch: main
- Last Pushed: 2026-06-10T23:07:43.000Z (about 1 month ago)
- Last Synced: 2026-06-11T00:19:48.975Z (about 1 month ago)
- Language: Python
- Size: 16.6 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# AgentStack
Starting out on a 6-part Medium Blog series on agentic system design for senior engineers.
Each article will have a standalone, fully runnable project. No shared dependencies between articles — clone the folder you need, install, run.
Mostly with **Google Antigravity SDK** and **Gemini 3.5 Flash model**.
By [Deepa Subramanian](https://www.linkedin.com/in/sdkdeepa/) — Google Developer Expert, Web Technologies
---
## Stack
| Tool | Purpose |
|------|---------|
| [Google Antigravity SDK](https://antigravity.google) | Agent loop, tool execution, context management |
| [Antigravity IDE](https://antigravity.google) | Development environment (VS Code-compatible, Gemini built-in) |
| Gemini 3.5 Flash | LLM — 1M token context window, strong function calling |
| Python 3.12 | All code sample|
---
## Quickstart
Every project follows the same three steps:
```bash
# 1. Go to the article folder
cd 01-token-budget-agent
# 2. Install dependencies
pip install -r requirements.txt
# 3. Set your API key
cp .env.example .env
# Edit .env — add your GEMINI_API_KEY
# 4. Run
python agent.py
```
**Get a free Gemini API key** at [aistudio.google.com](https://aistudio.google.com).
New Google Cloud accounts get **$300 free credit** — enough to run all 6 projects at scale.
---
## Why this series
Most agent tutorials show you a working demo and call it done. They skip the part where it fails.
This series covers the failure modes — memory degradation, context overflow, production gaps, multi-agent coordination failures, and the observability layer that most teams ship without. Each article is written from the perspective of someone who has shipped agents to production, not someone who has read about it.
If you have already seen the [Multi-Agent Code Review Pipeline](https://github.com/sdkdeepa/mulitagent-demo), this series is the next layer down. That project demonstrates what multi-agent coordination looks like. This series explains the architectural decisions underneath it.
---
## Subscribe to my technical writings
- Newsletter: [AI Architecture Weekly](https://bit.ly/aiarchitectureweekly) — for Engineering leaders
- Medium: [Blogs](https://medium.com/@sdkdeepa) - for Senior Software engineers