neochaotic

neochaotic / leoflow

Public

GitOps-first, container-native workflow orchestrator in Go (Airflow-UI compatible).

16
0
100% credibility
Found May 23, 2026 at 16 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Go
AI Summary

Leoflow is an open-source workflow orchestrator written in Go that replaces Apache Airflow's slow Python control plane. Users define pipelines in Python (similar to Airflow), then compile them into self-contained container images that run as ephemeral pods in Kubernetes. Each task gets its own clean environment, eliminating the dependency conflicts and worker leaks that plague traditional orchestrators. The system keeps the familiar Airflow vocabulary and UI, so teams migrate without retraining. The scheduler runs decisions in under 200 milliseconds (vs Airflow's 3-10 seconds), and the container-native model handles 100,000+ concurrent sensors efficiently.

How It Works

1
💡 You have a data pipeline problem

You've been fighting with Airflow: slow task starts, memory leaks, endless dependency conflicts. You hear about a tool that fixes all of it.

2
📝 You write your DAG in plain Python

Just like Airflow, you define tasks in a dag.py file using familiar decorators. You also add a simple leoflow.yaml file listing your dependencies like pandas and requests.

3
⚡ You compile your project

You run leoflow compile and watch as your project transforms into a container image. No Dockerfile to write, no Docker knowledge needed. Your pipeline becomes a portable artifact.

4
🚀 Everything deploys automatically

Your compiled DAG reaches the control plane. The scheduler keeps it running on its cron schedule, and each task spins up in its own clean environment, completely isolated from every other job.

5
👀 You watch everything in the familiar UI

The Airflow UI shows your DAG runs, task states, and logs exactly as you're used to. No learning a new dashboard, no teaching your team new vocabulary.

🎉 Your pipeline runs fast and reliable

Tasks start in under a second instead of minutes. Each run is clean and isolated. When something fails, retries happen automatically. You stop dreading 3am alerts.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 16 to 16 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is leoflow?

Leoflow is a workflow orchestrator written in Go that compiles Python DAGs into immutable container images and dispatches each task as an ephemeral Kubernetes pod. The control plane handles scheduling, state management, and log streaming over gRPC, with Postgres storing metadata and Redis holding task data (XCom values). It embeds the Apache Airflow 3.2.x UI so teams keep their familiar tooling without migration. A CLI (`leoflow init`, `leoflow compile`, `leoflow push`) handles the full authoring-to-deployment lifecycle.

Why is it gaining traction?

The pitch targets Airflow's five chronic wounds: scheduler stalls (3-10 seconds per decision), sensor concurrency limits (~500), repeated DAG reparsing, leaky Celery workers, and dependency hell across DAGs. Leoflow replaces the Python control plane with Go (no GIL), uses containers for isolation, and pre-compiles DAGs. The Airflow UI compatibility means no retraining. Scheduler decisions target sub-200ms latency versus Airflow's multi-second overhead.

Who should use this?

Data engineers running Airflow on Kubernetes who hit scheduler latency or worker memory leaks. Python developers comfortable with Airflow vocabulary who want container-native isolation without abandoning the UI. Teams managing DAGs with conflicting dependency versions. Not ready for teams needing production stability or vendor support.

Verdict

Leoflow has a compelling architecture solving real Airflow pain points, but 16 stars and pre-alpha status put it firmly in "early experiment" territory. The 1.0% credibility score reflects the project's infancy: limited community validation, no production deployments, and an aggressive feature roadmap still in progress. Watch it closely if you want container-native orchestration without Airflow's baggage, but wait for v0.1.0 before considering it for anything mission-critical.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.